Communications, Signal Processing, and Systems

Length and width feature of ship target is usually used as the initial criterion for ship type. A length and width feature extraction method of ship target based on IR image is proposed in this paper. At first, the preprocesses such as denoise and contrast enhancement are carried out, then the Hough transform is employed to detect the sea-sky-line, and the target potential area is determined, then edge detection and expansion and hole filling are used to obtain the whole connected region of the target. Finally, the minimum enclosing rectangle of the connected region is obtained according to the minimum area criterion, and the length and width of the minimum enclosing rectangle is the length and width of the ship target. The experimental results show that the method can effectively extract the length and width feature of ship target in complex sea-sky background, then with other auxiliary information can realize ship target recognition.

The contrast of lunar images is low, and few features can be extracted. Therefore, lunar images can be hardly matched with high accuracy. A lunar image matching method based on features from accelerated segment test (FAST) feature and speeded-up robust features (SURFs) descriptor is presented. First, entropy of image is adopted to automatically compute threshold for extracting FAST features. Second, SURF descriptors are used to describe candidate features, and then initial matches with nearest neighborhood strategy are obtained. Third, outliers are rejected from initial matches by RANSAC-based model estimation strategy and homography constraint. Experimental results show that the proposed method can get enough image correspondences and the matching errors are less than 0.2 pixels. It indicates that the proposed method can automatically achieve high-accuracy lunar image matching and lay good foundation for subsequent lunar image stitching and fusion.

This paper proposes a novel method to detect the water level of a river or reservoir. Images of the ruler which is used to measure the water level are obtained easily from a camera installed on the bank. Based on the property of the images captured by the camera, the problem of water level calculation can be transformed to the problem of classifying each image into two classes of ruler and water. As dictionary learning model has shown, its ability and efficiency in image classification problems, it is utilized in this paper to solve the problem of water level detection.

Aiming at the problems of fuzzy c-means clustering (FCM) and its improved algorithms for MRI image segmentation, this paper proposes a new FCM algorithm based on neighborhood pixel correlation. The algorithm works out the influence degree of the neighborhood pixels on the central pixel by the correlation of the gray-level difference between the domain pixel and the center pixel. Then, the distance between the neighborhood pixel and the cluster center is used to control the membership of the center pixel, the improved algorithm will solve the existing influence factors of unification, ignoring the difference between pixels, resulting in inaccuracy of segmentation results. At last, this algorithm is implemented by MATLAB tool and compared with FCMS and FLICM algorithms. The feasibility of the presented algorithm and the accuracy of the segmentation result are verified by evaluating the algorithm and the experimental results according to the relevant evaluation criteria.

Person re-identification (re-ID) is mainly used to search the target pedestrian in different cameras. In this paper, we employ generative adversarial network (GAN) to expand training samples and evaluate the performance of two different label assignment strategies for the generated samples. We also investigate how the number of generated samples influences the re-ID performance. We do several experiments on the Market1501 database, and the experimental results are of essential reference value to this research field.

With the continuous development of the social economy, the problem of low efficiency of invoice reimbursement has received more and more attention from companies, universities, and governments in China. In this paper, based on the recognition of invoices by OCR, we use Hough transform to preprocess the scanned image of invoices and creatively introduce the idea of checking the amount of money. We proofread the uppercase and lowercase amounts in the OCR recognition results. Using this method, the accuracy rate of OCR recognition increased from 95 to 99%, which greatly reduced the employees’ reimbursement time.

Since the theory of generative adversarial nets (GANs) put forward in 2014, various applications based on GANs have been developed. Most of the applications focused on generator network (G) of GANs to solve the daily challenges. However, rare of them had been aware of the great value of the discriminator network (D). In this paper, we propose a new method of detecting white Gaussian noise and its degree by the discriminator of generative adversarial nets. The results of our experiments show the feasibility of detecting white Gaussian noise (WGN) and evaluating its degree through generative adversarial nets.

Convolutional neural networks (CNN) have been applied to many classic problems in computer vision. This paper utilized CNNs to reconstruct super-resolution images from low-resolution images. To improve the performance of our model, four optimizations were added in the training process. After comparing the models with these four optimizations, Adam and RMSProp were found to achieve the optimal performance in peak signal to noise ratio (PSNR) and structural similarity index (SSIM). Considering both reconstruction accuracy and training speed, simulation results suggest that RMSProp optimization in the most scenarios.

It is important and necessary to obtain high-frequency information and texture details in the image reconstruction applications, such as image super-resolution. Hence, it is proposed the multi-scale fusion network (MCFN) in this paper. In the network, three pathways are designed for different receptive fields and scales, which are expected to obtain more texture details. Meanwhile, the local and global residual learning strategies are employed to prevent overfitting and to improve reconstruction quality. Compared with the classic convolutional neural network-based algorithms, the proposed method achieves better numerical and visual effects.

Due to the indefinite position of the characters in the invoice and the difference of the color shades, which greatly increases the difficulty of intelligent identification and thus it is difficult to meet practical applications. In order to solve this problem, this paper proposes a quadratic segmentation algorithm based on image enhancement. Specifically, we firstly enhance the color of the image based on gamma transformation and then separate the machine-printing character from the blank invoice based on the color analysis of the machine-printing character. Then according to the open operation in the image processing field and the boundingRect algorithm, the pixel information of the machine-playing character is obtained, which is convenient for getting the character information. The algorithm can achieve effective extraction of machine-playing characters and also reduce the difficulty of invoice identification and improving the accuracy of invoice identification. Simulation results are given to confirm the proposed algorithm.

In this paper, we proposed a real-time automated vehicle color recognition method using you look only once (YOLO)9000 object detection for intelligent transportation system applications in smart city. The workflow in our method contains only one step which achieves recognize vehicle colors from original images. The model proposed is trained and fine tuned for vehicle localization and color recognition so that it can be robust under different conditions (e.g., variations in background and lighting). Targeting a more realistic scenario, we introduce a dataset, called VDCR dataset, which collected on access surveillance. This dataset is comprised up of 5216 original images which include ten common colors of vehicles (white, black, red, blue, gray, golden, brown, green, yellow, and orange). In our proposed dataset, our method achieved the recognition rate of 95.47% and test-time for one image is 74.46 ms.

The single-pixel camera is a new architecture of camera proposed in recent years. The difference between a traditional camera and a single-pixel camera is that one image can be reconstructed by acquiring less amount of data with the latter. Most existing single-pixel cameras only collect data for one light path. In this paper, in order to reduce the impact of measurement noise, we adopt a way of dual-fiber acquisition to collect data. We compared the result of traditional single-fiber acquisition with our proposed dual-fibers acquisition. For video reconstruction, we use a dual-scale matrix as the image measurement matrix which can restore images with two different spatial resolutions as needed. We use the low-resolution video as a preview to acquire optical flow, and then we reconstruct a better-quality video by using the optical flow as a restrictive condition. We built an actual single-pixel camera hardware platform based on dual-fiber acquisition, and we show that our high-quality video can be restored by collecting data from our single-pixel camera.

Edge is in the place where image gray scale changes severely, it contains abundant image information. Image edge detection is a hot and difficult research field. Compare and analyze several classic edge detection method, aim at the advantages and disadvantages, respectively, propose a multi-scale edge detection algorithm based on the wavelet. The simulation shows the algorithm obtains an ideal effect in edge location and noise suppression.

The traditional Harris corner detection algorithm is sensitive to noise, and Corner is prone to drift at different image resolution. Combined with the multi-scale features of wavelet transform, propose a corner detection algorithm based on the wavelet transform. The algorithm maintains the advantages of Harris corner detection algorithm in image scaling, rotation or gray scale change, improves its disadvantage of scale invariance, and has strong anti-noise and real-time performance. It has good anti-noise and real-time performance.

Subject to the response uniformity of photoelectric sensors, the captured raw images always have serious chroma distortions. How to determine the mapping matrix between RGB and XYZ color spaces is important for the color distortion correction. However, the commonly used algorithms cannot give consideration to the precision and the adaptability. A more reasonable mapping algorithm based on variable-exponent polynomial regression is proposed to evaluate the mapping matrix coefficients. Variable-exponent regularization with the Lρ-norm (1 < ρ < 2) combines the features of lasso regression and ridge regression methods, owning both the sparsity and smoothing properties. The optimal solution for the variable-exponent regularization is given using lagged fix-point iteration method. Data from the standard color correction experiments are used to test the variable-exponent, lasso, ridge, and least-squares regression algorithms with different polynomial regression models. The results demonstrate that the proposed algorithm has the best performance.

Single-image super-resolution has been broadly applied in many fields such as military term, medical imaging, etc. In this paper, we mainly focus on the researches of recent years and classify them into non-deep learning SR algorithms and deep learning SR algorithms. For each classification, the basic concepts and algorithm processes are introduced. Furthermore, the paper discusses the advantages and disadvantages of different algorithms, which will offer potential research direction for the future development of SR.

Infrared image generation technology is an important part of infrared target guidance simulation. This paper studies the method of environment modeling, introduces the method of constructing SNR model and the method of coordinate transformation. The infrared image sequence is generated by SNR, coordinate information, and projectile information. The infrared image generated has three different target models: sky, sea, and land. The background is pure, the target is clear, and the experiment shows that it is feasible.

In this paper, non-sub-sampled shearlet transform (NSST) multi-scale analysis is combined with phase stretch transform (PST) to nonlinearly enhance images. The components of different scales after NSST multi-scale decomposition are processed by nonlinear models with different thresholds, and the noise is well suppressed while enhancing the detail features. The thresholds of the enhanced model are determined by the local standard deviation of PST feature map. Experiments on Matlab platform show that the proposed algorithm has improved image distortion, cleared details, and enhanced image contrast.

The convolutional neural networks have achieved very good results in the field of remote sensing image classification and recognition. However, the cost of huge computational complexity with the significant accuracy improvement of CNNs makes a huge challenge to hardware implementation. A promising solution is FPGA due to it supports parallel computing with low power consumption. In this paper, LeNet-5-based remote sensing image classification method is implemented on FPGA. The test images with a size of 126 × 126 are transformed to the system from PC by serial port. The classification accuracy is 98.18% tested on the designed system, which is the same as that on PC. In the term of efficiency, the designed system runs 2.29 ms per image, which satisfies the real-time requirements.

Feature extraction and classification technology based on hyperspectral data have been a hot issue. Recently, the convolutional neural network (CNN) has attracted more attention in the field of hyperspectral image classification. To enhance the feature extracted from the hidden layers, in this paper a deconvolution layer is introduced in the deep 2DCNN model. Analyzing the function of convolution and pooling to determine the structure of the convolutional neural network, deconvolution is used to map low-dimensional features into high-dimensional input; the target pixel and its pixels in a certain neighborhood are input into the network as input data. Experiments on two public available hyperspectral data sets show that the deconvolution layer can better generalize features for the hyperspectral image and the proposed 2DCNN classification method can effectively improve the classification accuracy in comparison with other feature extraction methods.

Micro-video has fragmented propagation mode and short, flat, and fast features. The convenience of its shooting and dissemination is widely used by the public. In addition to the development of Internet technology, the video compression technology has also played an important role. In this paper, three kinds of micro-video compression coding techniques are selected and analyzed; they are MPEG-4, AVC/H.264, and HEVC/H.265. The three video compression techniques are compared by the film trailer compression experiment, and the conclusion is applied to the micro-video compression of scientific research projects. Combining the actual advantages and disadvantages of the scheme, it is beneficial to the effective application of micro-video coding technology in practical projects.

The $$ L_{1/2} $$ regularization has been considered as a more effective relaxation method to approximate the optimal $$ L_{0} $$ sparse solution than $$ L_{1} $$ in CS. To improve the recovery performance of $$ L_{1/2} $$ regularization, this study proposes a multiple sub-wavelet-dictionaries-based adaptive iteratively weighted $$ L_{1/2} $$ regularization algorithm (called MUSAI- $$ L_{1/2} $$ ), and considering the key rule of the weighted parameter (or regularization parameter) in optimization progress, we propose the adaptive scheme for parameter $$ \lambda_{d} $$ to weight the regularization term which is a composition of the sub-dictionaries. Numerical experiments confirm that the proposed MUSAI- $$ L_{1/2} $$ can significantly improve the recovery performance than the previous works.

This paper conducted depth and systematic research on the influence of image motion on radiation quality. The exploration of the regular image radiation quality could be a great significance for improving the image quality. The source of abnormal image motion was analyzed in terms of platform motion and flutter, attitude control error, and image motion compensation error. The model of point spread function under abnormal image motion was established based on the analysis of superposition model between adjacent pixels and PSF variation model. The radiation information and MTF caused by PSF changes were analyzed. The experiment shows that the authenticity of the radiation information expressing the imaging target is weakened. The larger the abnormal image motion, the faster the MTF curve declines, indicates the overall image quality decreases, the spatial resolution decreases, and the sharpness decreases.

This paper proposed a flame segmentation algorithm based on the saliency of motion and color. First, feature point detection is performed on the video image using the scale-invariant feature transform (SIFT) algorithm, and the optical flow field of moving object in the adjacent frame is acquired by the optical flow method. According to the optical flow vector difference between the target pixel point and the surrounding neighborhood pixels, the motion saliency map is obtained based on the Munsell color system. Then, the LSI flame color statistical model based on the Lab and HSI space is used to extract color saliency map of video images. Finally, under the Bayes framework, the motion saliency map and the color saliency map are fused in an interactive manner to obtain the final flame segmentation map. Experimental results show that the proposed algorithm can effectively segment the flame image in different scenarios.

In this paper, a ring array imaging method for near-filed security screening is proposed for densely populated areas security imaging. With the fixed transceivers, this method can avoid the motion phase error, which is ineluctable in traditional synthetic aperture radar (SAR). Besides, the number of transceivers is less than planar array radar with the same quality of imaging results. The convolution back propagate (CBP) algorithm is adopted to reconstruct image to avoid the phase error, which is created by plane-wave hypothesis for near-field imaging in high-frequency band. Moreover, the simulation results of CBP and other imaging algorithm are described and compared. And the experimental results are shown to verify the performance of this method.

Road lighting is a main factor which impacts on traffic accident rate. The valuable lighting evaluations are the fundament of road lighting design. We propose five classes parameters which come from road lighting images to evaluate the quality of road lighting in this paper. We first calculate 10 image parameters from road lighting images. It includes mean value of gray level, variance of gray level, radiation precision steepness, gray level entropy, second moment of angle, contrast, autocorrelation, inverse difference moment, detail energy, and edge energy. Then, we divide the above 10 parameters into five categories using cluster analysis. These categories are mean value class, variance class, contrast class, detail energy class, and information-related class. Finally, combined with the physical meaning of the parameters, the evaluation index of the traditional road lighting and the characteristics of the human eye, we connect these five categories with the average brightness of pavement, the uniformity of road surface brightness, glare, road sign inducibility, and psychological factors. The experimental results show that the road lighting image parameters have good clustering properties, and the clustered image parameters can reflect the quality of road lighting.

As a typical feature point with the distinct advantage of being detected easily, the circle has been widely used for camera calibration and motion measurement. However, motion blur may cause a negative effect on the accuracy of the center location. In this paper, the developed method for the circle detection on motion blur image is proposed, which consists of two procedures. Wiener filtering is used to restore a degraded image in the first step. Zernike moment is utilized to subpixel central location in the second step. Image restoring simulation and center detection experiments are provided to verify the performance of the method. Results show that the clarity of the images restored by Weiner filtering is high and the circles on the restored image can be detected successfully and located accurately.

To solve the problem that image enhancement of finger vein patterns based on traditional filtering methods fails to intuitively highlight the feature of edge protection, the experimental study model based on the guided filter is proposed. Through adding the comparison experiment between guided filter and bilateral filter, and doing the binary processing to the finger vein image after the process of the guided filtering and bilateral filtering, it can be found that some noises exist around the vein texture. In order to reduce or eliminate the interference, a traditional average filtering method is applied for denoising, which not only highlights the vein texture details but eliminates the interference in the post-processing, and at the same time, adjusting the filter parameters will cause a significant impact on the enhancement of finger vein image. A comparison experiment in false recognition rate between two filtering algorithms is conducted, and visual and numerical evaluations are performed on finger vein image after the process of enhancement and binarization; the result indicates that the guided filter has better edge protection feature and lower false recognition rate than the bilateral filter.

In this paper, an improved LBF model based on local regional information is proposed for image segmentation. The basic idea is to add the regularized area term to the energy function of the LBF model and establish a variable coefficient with adaptive capability composed of image regional information. Compared with the LBF model, the proposed model increases the driving force of the evolution curve, making the result better when dealing with the images with weak boundaries and intensity inhomogeneity. At the same time, it effectively solves the problem that the LBF model is sensitive to the initial position and size of the evolution curve. This model is used to segment medical images with complex topological structure and intensity inhomogeneity. Experimental results show that regardless of the initial curve of any position or any size, it has little influence on the segmentation result; moreover, the localization of deep-depressed image boundaries is more accurate, so we get the conclusion that the new model has corresponding improvements in segmentation accuracy and robustness.

Currently, the image super-resolution reconstruction method based on sparse representation has limited ability to process the details of the edge. Therefore, based on the dictionary learning, the local variance feature edge gradient estimation image fast super-resolution reconstruction is improved and optimized based on dictionary training. The dictionary training process includes cluster analysis of high-resolution images, local variance extraction, and sparse filtering. The reconstruction process includes local variance detection of the low-resolution image and threshold judgment, and then the image is reconstructed according to the gradient value.

Satellite jitter, the fluctuation of satellite point, has negative influence on the geometric quality of high-resolution optical satellite imagery. Gaofen-1 02/03/04 satellites are Chinese operational remote sensing satellites, which have Earth observation capability with 2/8 (panchromatic/multispectral image) meter resolution. This paper presented a jitter detection method based on multispectral imagery for Gaofen-1 02/03/04 satellites. Three short strip datasets captured in the early in-flight period were used to conduct the experiments. The results indicate that satellite jitter with a frequency of 1.1–1.2 Hz, and amplitude of 1–2 pixel exists during the early in-flight period.

In order to improve the performance of fractal coding methods, a new method is proposed in this paper. Firstly, we find that the range blocks with large variances play a more important role in causing the degradation of decoded images, and the effect of the remaining range blocks can be ignored. Secondly, the range blocks with larger variances will be encoded in an extended domain block pool, and the remaining ones will be encoded with the no-search fractal encoding method. Finally, two fractal coding methods are used to assess the performance of the proposed method. Experiments show that compared with the previous methods, the proposed method can provide shorter encoding time, better quality of decoded images and fewer bits per pixel.

With the rapid development of communication technology and the Internet, the security of image information has also received more attention. The information characteristics of an image can be described by pixel position and pixel value. Therefore, the encryption algorithm for the image is generally based on these two aspects (Commun Nonlinear Sci Number Simulat 17(7):2969–2977, [1]; Int J Bi-furcation Chaos 7(7):1579–1997, [2]). Using a method alone can easily be cracked by an attacker, so image encryption often combines pixel position scrambling with gray value diffusion. In this paper, chaotic sequences are applied to image data encryption, and pixel position scrambling and gray value diffusion are combined to obtain ciphertext images. And the feedback mechanism is introduced in the diffusion link, so that the ciphertext of each pixel is not only related to the current plaintext, but also relevant to the ciphertext of all the previous pixels, thereby increasing the complexity of the ciphertext.

Weighted nuclear norm minimization (WNNM) uses image non-local similarity to deal with image denoising; this method not only maintains the detailed texture edge structure but also reduces the impact on distortion of the image after denoising. However, WNNM method assumes that the noise variance of the image is known, where the parameter is set by subjective experience that will result in incompleteness in theory. To handle this issue, it is proposed to pre-estimate noise variance based on discrete wavelet transformation (DWT). The simulation result shows that compared with original WNNM method, pre-estimate noise variance in image denoising has a faster algorithm running speed and a higher image signal-to-noise ratio after denoising.

The campus online behavior analysis needs to graphically display the distribution of the network traffic and the students’ application status in the campus network at present or in a period of time. After capturing and analyzing the campus network traffic data, statistics and analysis are performed on the acquired data. Based on the result, a visual display can be obtained. It helps the network administrator understand whether the use of campus network traffic is reasonable and can facilitate student management personnel to understand whether the students’ Internet behavior is healthy. In this paper, the B/S framework is used to capture and analyze the data obtained by the campus network using Web pages.

With the rapid development of the Internet of things, social networks, and e-commerce, the era of big data has arrived. Although big data has great potential for many areas such as industry, education, and health care, getting valuable knowledge from big data can be a daunting task. Big data has the characteristics of high-speed change, and its content and distribution characteristics are in dynamic changes. Most current models are static learning models that do not support online updating, making it difficult to learn dynamically changing big data features in real time. In order to solve this problem, this paper proposed a method to support incremental recursive least squares (IRLS) regression parameter estimation and variable sliding window algorithm to analyze and judge the trends of dynamic characteristics of data streams, which can provide early warning, status assessment, and decision support for monitoring objects and improve the accuracy and adaptability of data flow classification. The real-time computational and analysis accuracy are obviously improved than the traditional algorithm, and the simulation results verify the effectiveness of the proposed algorithm.

In order to make the civil aviation navigation system digital and miniaturization, a digital signal processing technology of applying the Doppler VHF Omnidirectional Range (DVOR) is proposed to improving the processing speed and the performance of the receiver data. This paper introduces the basic principle of DVOR signal and the basic principle and realization method of digital signal processing technology. Finally, we completed the design of digital signal demodulation, filtering, and comparison phase by LabVIEW 2016, so that the DVOR measure system can display information such as bearing.

Hydraulic pump vibration signal preprocessing is the basis for failure prediction. The vibration signal preprocessing method based on ILCD fusion is proposed to solve the problem that the vibration signal is nonlinear and the feature information is weak. Firstly, the high-frequency harmonics are combined with ILCD to decompose multi-channel vibration signals, and ISC components can be achieved. Secondly, the sensitive factors are defined as evaluation indexes, and the sensitive components containing the fault feature information are screened and weighted fusion. The reconstructed signal is obtained to reduce the noise and other interference components and effectively extract the fault feature information. Finally, the effectiveness of this method is verified by measuring the vibration signal of the hydraulic pump.

In this paper, we propose a real-time interactive method for mobile terminal users to access remote three-dimensional scene based on cloud services. We realize a dynamic and fluent method to access remote three-dimensional scene for mobile terminal. The method achieves a good user experience. We distribute mobile phone user task request and pad user task request from PC terminal user task request by the load balancing server and transfer these tasks to cloud server. So that mobile users transmit the interactive access operation in the scene to cloud server equivalent in the way of instruction set. We complete user command related work and high-quality rendering process by cloud server, which makes mobile terminal user access remote 3D scene immersive interaction. The mobile terminal user only needs to send a command request and receive high-quality pictures. The cloud servers receive instructions from user terminal and complete all related interactive work and transmit results in the form of high-quality pictures to mobile terminal user. This paper solves the technical challenge of mobile device users to access remote real-time interactive 3D scene in a manner of cloud services and continuous interval frame image for the first time.

The two-dimensional trajectory correction fuze of movable canard is the hot spot of research currently with the advantages of low cost and strong correction ability. For the two-dimensional trajectory correction fuze of movable rudder, continuous sine control, interval sinusoidal control, and constant control angle control are used to control the electric proportional canard. By analyzing the influence of different control methods on the ballistic characteristics and the correction ability, the canard control methods are evaluated and an optimal control method is proposed. It is greatly significant for the development of two-dimensional trajectory correction fuze for movable canard in theory and engineering.

For clutter with long tailing characteristics, such as Weibull distribution, Ruili distribution, lognormality, and K-distribution, CFAR processors corresponding to various distributions need to be used when CFAR processing technology is used. Otherwise, it is difficult to obtain constant false alarm characteristics when the statistical characteristics of clutter change. From the viewpoint of improving the robustness of CFAR, according to the central limit theorem and the logarithmic compression principle of the signal, this paper attempts to accumulate and average the radar signal before CFAR processing according to the pulse accumulation characteristics and median limit theorem of radar signals. The clutter of the post-PDF is close to a normal distribution, which effectively eliminates the trailing effect of the clutter characteristics, i.e., it effectively suppresses the sharp peak interference and the distribution characteristics of the normalized clutter.

Turbo codes are among the best error-correcting codes, but trade-offs between performance and complexity in decoding are required for hardware implementation. In this paper, a novel extrinsic information scaling scheme for max-log-MAP decoder is proposed. It scales and combines extrinsic information generated at successive iteration round. The proposed method is evaluated for 3GPP LTE turbo codes in terms of decoding performance, complexity, and convergence. The simulation results show it has decoding gain near to log-MAP while keeps almost the same computation complexity as max-log-MAP with slight increment in memory resource. Moreover, it maintains insensitivity to SNR estimation error of max-log-MAP algorithm. Compared with conventional scaling scheme, it accelerates extrinsic information exchange between two constituent decoders to get better convergence and decoding performance.

Current research on the classification for uncertain data mainly focuses on the structural changes of the classification algorithms. Existing methods have achieved encouraging results; however, they do not take an effective trade-off between accuracy and running time, and they do not have good portability. This paper proposed a new framework to solve the classification problem of uncertain data from data processing point. The proposed algorithm represents the distribution of raw data by a sampling method, which means that the uncertain data are converted into determined data. The proposed framework is suitable for all classifiers, and then, XGBoost is adopted as a specific classifier in this paper. The experimental results show that the proposed method is an effective way of handling the classification problem for uncertain data.

The detection of cycle slips and outliers in single-frequency carrier phase data, or any other type of un-expected changes in the single-frequency carrier phase measurements of the GNSS, is one of the major data preprocessing problems that needs to be addressed, especially single-frequency receivers account for most of the market share and GNSS carrier phase data are used for real-time applications that require reliable position results. In this contribution the improved RKF (Robust Kalman Filter) is designed to detect cycle slips or unexpected changes when single-frequency carrier phase is interfered by small outliers, in order to improve the cycle slip detection success rates. Real BDS single-frequency data have been used to test and evaluate the algorithm, where the simulation results indicate that the improved RKF has a higher cycle slip detection success rate than the RKF when observations are interfered by small outliers, proving the efficiency of the algorithm proposed in this paper.

Regenerating codes (RGCs) are considered as optimal trade-off between repair bandwidth and storage amount per node in a distributed storage system (DSS). Actually, due to the limited nodes’ amount and bandwidth resources in networks, redundant bandwidth cost and assisted nodes connections will involve as employing the traditional RGC. Specific to these problems, we propose a new code, named shorten minimum storage regenerating code (sMSR) with two novel targets, unit storage cost (USC) and unit repair bandwidth (URB), and construct it by removing some information bits of the mother code generated by product matrix in encoding process. Additionally, in order to improve the availability of sMSR, we implement Binary Addition and Shift Implementable Convolution (BASIC) to decrease the computation complexity. The simulation results demonstrate that our code improves repair efficiency of MSR codes in practical DSS.

Background subtraction is an important task in computer vision. Pixel-based methods have a high processing speed and low complexity. But when the video frame with camouflage problem is processed, this kind of methods usually output incomplete foreground. In addition, the parameters of many algorithms are invariable. These methods cannot tackle non-static background. In this paper, we present an adaptive background subtraction algorithm derived from ViBe. Gaussian Kernel template is used to model initialization and update. Standard deviation is used to measure background dynamics. We test our algorithm on a public dataset, named changedetection.net. The results show that we can handle most of scenarios. Compared to ViBe, we achieve better result generally, especially in dynamic background and camera jitter categories.

Through the research and analysis of a relatively novel natural heuristic, meta-heuristic swarm intelligence optimization algorithm, this swarm intelligence algorithm is defined as a whale optimization algorithm. The algorithm builds a mathematical model by simulating a social behavior of humpback whales. This optimization algorithm was inspired by the bubble-like net hunting phenomenon that humpback whales prey on. By analyzing the four benchmark optimization problems with or without offset and rotation, the convergence performance of the whale optimization algorithm and the ability to solve the optimization problem are proved. The performance of the whale optimization algorithm is based on the computer simulation technology. Through the convergence curve obtained from the experiment, we can see that the whale optimization algorithm performs best for the five benchmark optimization problems without rotation.

With the rapid development of information technology, biometric technology has been widely used. However, biological information is limited to everyone, once it is leaked, it can no longer be safely used, which will cause lifelong damage to users. Therefore, the study of biological template protection technology is of great significance. This paper introduces the deoxyribonucleic acid (DNA) coding theory into biometric template protection and proposes a face template protection algorithm based on DNA coding encryption. Experiments and analyses are carried out on Olivetti Research Laboratory (ORL) face database. The results show that the proposed algorithm not only enhances the security of the original template, but also ensures the recognition performance of the system.

As an indicator of emotion intensity, arousal plays an important role in emotion recognition. However, the accuracy rate-based EEG signals have been far away from human’s satisfactory due to the lack of effective methods. In this paper, we propose a novel framework for recognizing arousal levels by using EEG signals. Instead of using time domain feature and frequency domain feature of EEG, we select the EEG feature directly from a large number of EEG signals by using the feature selection method after data standardization. Based on our method, feature with most distinguished ability has been found. The experimental results on the open data set DEAP show that the arousal accuracy has been significantly improved by using our method.

In order to improve unequal error protection (UEP) for compressed images, a novel rateless transmitting scheme is proposed in this paper. Based on 3GPP Raptor codes, this scheme can accurately match the image with different priority to ensure the efficiency and reliability. These Raptor codes have the linear complexity of encoding and decoding. In addition, a new framework called Short Interweaving Systematic Raptor (SISR) codes is proposed. Based on Markov process, the theoretical analysis of degree distribution design is given. Simulation results show that, compared to the traditional Raptor code, SISR can prominently improve the PSNR to obtain better image quality with different packet loss rate.

Diffusion adaptive networks are received attractive applications in various fields such as wireless communications. Selections of combination policies greatly influence the performance of diffusion adaptive networks. Many diffusion combination policies have been developed for the diffusion adaptive networks. However, these methods are focused either on steady-state mean square performance or on convergence speed. This paper proposes an effective combination policy, which is named as relative-deviation combination policy and uses the Euclidean norm of instantaneous deviation between the intermediate estimation vector of alone agent and the fused estimation weight to determine the combination weights of each neighbor. Computer simulations verify that the proposed combination policy outperforms the existing combination rules either in steady-state error or in convergence rate under various signal-to-noise ratio (SNR) environments.

With the promotion of “the Belt and Road” and twenty-first-century “Maritime Silk Road” and the establishment of an information network in Hainan Province, the construction of network covering Southern China Sea is particularly important, which should combine with cable, wireless and satellite transmissions. The complexity of the impact of the marine environment on radio wave propagation and the particularity of communication between vessels are important factors to be considered when designing marine radio communication systems. This paper studies the models of sea channel for sea surface diffuse reflection. In this model, the performance of OFDM transmission was analyzed.

Digital signal processing on high-performance computing platform based on general-purpose processor is one of the development directions of software-defined radio, which provides a new idea for the signal combining of large-scale antenna array. In this paper, two parallel implementations of SUMPLE combined algorithm are proposed, which are multi-antenna-group method (MAGM) based on task decomposition principle and multi-signal-block method (MSBM) based on data decomposition principle. The processing capability and real-time performance of the two methods are analyzed in detail. Results show that MAGM is not suitable for the high-performance computing platform because of large and frequent data transmissions from central processing unit (CPU) to graphics processing unit (GPU). The traditional iterative method of SUMPLE cannot be applied to MSBM due to the dependencies among signal blocks.

Timing recovery is one of the most important issues that need to be solved for a receiver. The key of timing recovery is the extraction of timing errors. Currently, the timing recovery in the satellite communication systems cannot meet the requirement of the real-time passive intermodulation (PIM) interference suppression. It calls for the study on improved algorithm that can be applied in the satellite communication system. And the improved algorithm should have small computation and a high running speed. This paper first introduces the overall structure of the traditional square timing recovery loop, and then proposes an improved square timing error detection algorithm. By using the Goertezl algorithm, this improved algorithm optimizes the step–extracting spectral components at the symbol rate 1/T. Finally, this paper simulates and analyzes the improved algorithm, as well as parameters’ influence on the performance of the algorithm. It is proved that the improved square timing error detection algorithm can greatly reduce the amount of computation under the premise of guaranteeing performance.

In the future, information service calls for higher requirements for space-based information network infrastructure. However, there are problems of independent development and redundant construction among existing space-based information systems. To solve the problems, a software-defined space-based integration network architecture is proposed in this article, which could provide information transmission, networked awareness, position navigation and timing (PNT) service on demand.

To solve the problem of complex maneuvering target tracking, an improved interacting multiple model (IS-AR-IMM) algorithm based on AR model set switching is proposed in this paper. The algorithm is proposed on the basis of the AR-IMM algorithm based on AR model and S-IMM algorithm based on model set switching. Simulation shows that IS-AR-IMM algorithm not only solves the problem of increasing computational complexity and model selection caused by the increase of models in AR-IMM and S-IMM algorithms, but also has better tracking performance.

For 3D reconstruction algorithm based on binocular stereo vision only approximately restores the 3D shape of the object, but it does not analyze the performance of using 3D reconstruction algorithm based on the model of binocular stereo vision measuring distance device to restore object, an interval constraint range domain algorithm is proposed in this paper. In this paper, this algorithm finds the boundary points of the object, the performance of using 3D reconstruction algorithm to restore object depends on the proportion of the number of points that fall within the constraint interval to the total number of points. The interval constraint range domain algorithm analyzes the performance of the 3D reconstruction algorithm to recover the object, that is, analyzes the measurement accuracy of the algorithm. Under the condition of the resolution of the image sensor used in the algorithm is identical, the simulation results show that the accuracy of the algorithm is determined by the spacing of the image sensor, the position, and shape of the constraint interval.

The distributed compressive video sensing (DCVS) system greatly reduces the pressure on the encoder by transferring the computational complexity to the decoder, which is suitable for the limited-resource video sensing and transmission environment, in the meantime, get the better performance from key (K) frames and non-key (CS) frames. In this paper, we use the approximate message passing (AMP) algorithm reconstruct the K-frames. In order to improve the quality of the reconstructed K-frames, we add the block-matching 3D filtering (BM3D) denoising strategy based on the AMP algorithm. For the CS-frames, we improve the reconstructed CS-frames by improving the accuracy of side information (SI) frames by proposing a new high efficiency video coding (HEVC) motion estimation (ME) algorithm with motion vector (MV) prediction method. After we obtain the SI frames and combine the compressed value of the CS-frames with the side information (SI) fusion algorithm based on the difference compensation algorithm, the high accuracy SI frame is integrated into the reconstruction algorithm of the CS-frames. The experimental results demonstrate that our algorithms achieve higher subjective visual quality and peak signal-to-noise ratio than the traditional methods.

Based on the analysis of the existing consensus mechanism of blockchain, this paper proposes a POWS mechanism based on the adjustment of workload. In the POW consensus mechanism, the concept of coinage is introduced to adjust the mining difficulty of different nodes. The POWS adjusts the mining difficulty through two factors: calculation force and coinage. The POWS is compared with the POW and the POS in two aspects, the basic performance and the ability to resist the mining pool. The experimental results show that the POWS consensus mechanism can meet the performance requirements of normal blockchain system, at the same time it can better diminish the impact of calculation force and coinage on block generate efficiency, narrow the efficiency gap between the mining pool nodes and the non-mineral pool nodes, and reduce the interest of the mining pool to non-mineral pool nodes.

Karst aquifers supply drinking water for 25% of population on the earth. Better understanding of the heterogeneity of karst aquifers can help us develop karst groundwater sustainably. Karst hydrological processes include precipitation (rainfall) infiltration, groundwater wave propagation in karst aquifers, and spring discharge. The processes of precipitation signals’ transformation into spring discharge signal are mainly affected by heterogeneity of karst aquifers. Analysis of relations between spring discharge and precipitation can identify heterogeneity of karst aquifers. This paper explores the periodic characteristics of spring discharge and precipitation, and the heterogeneity of karst aquifers in Niangziguan Springs (NS) using multitaper method (MTM). The results show that both spring discharge and precipitation exist in the same period of one year. Cross-correlation function is used to calculate the time lags between the reconstructed spring discharge and precipitation in different areas of the NS basin. The results indicate that the response time of the spring discharge to precipitation is different at different areas. The time lag between the spring discharge and precipitation is 3 months at Pingding County; 4 months at Yu County, Yangquan City, Xiyang County, and Heshun County; and 27 months at Shouyang County and Zuoquan County. The results reflect the heterogeneity of the NS basin and are consistent with the geological structure of the NS basin. MTM is robust in identification of heterogeneity of karst aquifers.

The correlations between spectrum state evolutions, as a kind of similarity measure, have been revealed to optimize the spectrum usage model or improve the performance in spectrum prediction. However, most existing similarity analyses only end up with the superficial similarity phenomenon. It is of great need for us to conduct the deep investigation and analysis on the similarity of spectrum state evolutions. Firstly, we design a similarity index for spectrum state evolutions based on the Euclidean distance. Then, a network of spectrum state evolutions in the frequency domain can be formed for further analysis by comparing the proposed similarity indexes of frequency points with the decision threshold. Experiments with real-world spectrum data prove the feasibility and rationality of the above similarity analysis.

The carrier frequency offset is a common problem in communication systems, but in most case, phase-locked loop (PLL) can only track small frequency shift, so an estimation of initial frequency offset is necessary for CPM synchronization. In coded multi-h CPM system, it is a convenient method to use the information of frame header for frequency estimation. In this paper, we combine the pulse-pair technique and fast Fourier transform (FFT) calculating method for our purpose. According to the simulation results, it can achieve accurate estimation for comparative great frequency offset and perform steady even in low signal-to-noise ratios (SNRs).

Based on the definition of relative entropy and the principle of minimum relative entropy, this paper takes the pressure of the node as the main research object and establishes an optimization model about the water pressure of the pipe network running in the water supply pump station. Then we apply genetic algorithm to solve our model and analyze the water pressure of the actual water supply pumping station in FS city. According to the model established in this paper, we can optimize the relevant decision variables in the water distribution network and provide a new method basis for the explicit pump scheduling.

Emerging applications, such as mobile cloud computing and WSN, bring lots of challenges to current data center networks, and many of them need to construct their own virtual networks (VNs) for the flow-line calculation. Besides, delay-sensitive applications, like IoV and mobile 5G broadband application, require to move the computing resource and some parts of databases closed to customer sites. Thus, the computing resource near application fields, the so-called edge computing, becomes more and more in short supply; meanwhile, the backbone network is under an increasing pressure on the long-haul communication workload. In this paper, we first make a knowledge-aware of VN and design a mapping orchestration based on the QoS requirement features and workload matching. Here, we distinguish the computing requirement from each virtual node in VN and allocate them into edge or cloud DC, respectively. Then, we formulate this resource allocation problem and propose a fast and efficient algorithm. Finally, the numerical results verified the advantages of our algorithm in terms of average computing latency and average transmission latency.

Generalized frequency division multiplexing (GFDM) is a flexible multicarrier modulation scheme and has been considered as a candidate for the physical layer of the 5G mobile communication system. The bit error rate performance of GFDM is affected by the symbol timing synchronization errors; however, the existing preamble-based symbol timing synchronization scheme for GFDM systems suffers from the timing-ambiguous plateau problem. In this paper, two improved preamble structures are constructed according to the characteristics of GFDM symbols to eliminate the timing-ambiguous plateau. The simulation results under AWGN and multipath fading channels show that the proposed autocorrelation timing synchronization schemes have better timing MSE performance compared to the existing preamble-based timing synchronization schemes.

Maritime wireless communication is in the stage of continuous development. Medium and short wave communication has always been the main method of maritime communications. Realizing rapid, efficient, and reliable signal transmission is the urgent need for the development of current situation. In this paper, aiming at MF, HF, or VHF radio waves emitted by land, Bayesian linear regression model is used to solve the reflection problem of calm sea surface, and then EM (expectation maximization) algorithm is further used to solve the complex Bayesian model to solve the reflection problem of rough sea surface. Furthermore, the advantages of the Bayesian linear regression model over other models, such as the Longley-Rice model, are obtained.

Target classification in foliage environment is a challenging task in realistic due to the high-clutter background and unsettled weather. To detect a particular target, e.g., human, under such an environment, is an indispensable technique with significant application value. Traditional method such as computer vision techniques is hardly leveraged since the working condition is limited. Therefore, in this paper, we attempt to tackle human detection by using the radio frequency (RF) signal with a device-free sensing. To this end, we propose a differential evolution flower pollination algorithm support vector machine (DEFPA-SVM) approach to detect human among other targets, e.g., iron cupboard and wooden board. This task can be formally described as a target classification problem. In our experiment, the proposed DEFPA-SVM can effectively attain the best performance compared to other classical multi-target classification models and achieve a faster convergent speed than the traditional FPA-SVM.

In this paper, we first consider the random caching and carefully analyze the signal-to-interference-plus-noise ratio from the perspective of a typical user, deriving the expression of successful transmission probability. Then, our objective is to maximize the successful transmission probability by optimizing the caching distribution in high SNR using proposed subgradient descent method. Finally, simulations illustrate that the caching distribution optimized by our proposed algorithm achieves a gain in successful transmission probability over existing caching distributions.

The concentration of nitrogen and phosphorus in the waters is an important indicator to affect water quality and determine the degree of water pollution. The development of hyperspectral remote sensing technology makes it possible to monitor the concentrations of nitrogen and phosphorus in different water areas. In this experiment, the spectral curves of different concentrations of nitrogen and phosphorus solutions were collected by using an imaging spectrometer under laboratory conditions. Then compare the spectral curves of different concentrations of sodium phosphate solution by PSR spectrometer, to analyze the sensitive bands of nitrogen and phosphorus. The experimental results show that for phosphorus, its concentration as a whole is positively correlated with the spectral reflectance. In the wavelength range of 450–630 nm, there is a strong positive correlation between the concentration of phosphorus and the spectral reflectance. The correlation coefficient is above 0.8, and the maximum positive correlation is 0.9 at 550, 603, and 740 nm. For nitrogen, its concentration as a whole is negatively correlated with the spectral reflectance. In the wavelength range of 560–850 nm, the correlation coefficient fluctuates within the range from −0.6 to 0.95, and the maximum negative correlation of 0.95 is achieved at 603, 670, and 807 nm.

With the increase of aerospace launch density, the stability of the firing range measurement and control system and the network communication system in the range is particularly important. The potential failure of the range information system needs more attention, including the aging of the line, the destruction of animals, human damage, and the influence of viruses. Electromagnetic interference, etc., may cause serious problems such as delay in launching missions, errors in receiving and monitoring signals, and inability to issue satellite in-orbit control commands, even causing major accidents involving star destruction. In order to adapt to the load capacity during the high-density task period, to enhance the cognitive ability of the new load launch, and to improve the ability of the range to perform difficult tasks, it is necessary to accurately diagnose and maintain the launch system of the space range.

Anomaly detection becomes increasingly important in hyperspectral data exploitation due to the use of high spectral resolution to uncover many unknown substances which cannot be visualized or known a priori. The RX detector is one of the most commonly used anomaly detections algorithms, where both the global and local versions are studied. In the double window model of local RX detection, it is inevitable that there will be abnormal pixels in the outer window where the background information is estimated. These abnormal pixels will cause great interference to the detection result. Aiming at a better estimation of the local background, a joint kurtosis–skewness algorithm is proposed to smooth the background and get better detection results. The skewness and kurtosis are three and four order statistics respectively, which can express the non-Gaussian character of hyperspectral image and highlight the abnormal information of the target. The experimental results show that the proposed detection algorithm is more effective for both synthetic and real hyperspectral images.

Sub-band segmentation and reconstruction technology is the core technology of the antenna array signal full spectrum combine scheme. Based on the principle of complex exponential modulation filter banks, the sub-band segmentation and reconstruction technique based on the Weighted OverLap-Add-(WOLA)-structured filter bank is studied. A filter bank with a sub-band number of 256 and an oversampling factor of 1.45 are designed. Compared with the multiphase DFT-structured filter banks currently used, the filter can not only realize the segmentation and reconstruction of performance and considerable, but also break through the oversampling factor must be integer constraints, the structure is more efficient and more flexible parameter settings.

Aiming at detecting cracks in photovoltaic images, a crack detection algorithm of photovoltaic images based on Laplacian pyramid decomposition is studied in this paper. Firstly, in order to suppress noise from the crack area, the image is subjected to a filtering process and contrast enhancement operation. Then, the multi-scale edge detection based on Laplacian pyramid decomposition is applied to the processed image to extract the edge of the image. The results of the extracted fractures are optimized to eliminate the influence of partial noise. Through tests and comparisons, the algorithm is proved effective on crack detection for photovoltaic image.

In this paper, we proposed a method to synthesizing a super-resolution image with the given image and text descriptions. Our work contains two parts. Wasserstein GAN is used to generate low-level resolution image under the guidance of a novel loss function. Then, a convolution net is followed to refine the resolution. This is an end-to-end network architecture. We have validated our model on Caltech-200 bird dataset, Oxford-102 flower dataset, and BSD300 dataset. The experiments show that the generated images not only match the given descriptions well but also maintain detailed features of original images with a higher resolution.

As a famous heuristic algorithm, bat algorithm (BA) simulates the behavior of bat echolocation, which has simple model, fast convergence and distributed characteristics. But it also has some defects like slow convergence and low optimizing accuracy. Facing the shortages above, an optimization bat algorithm based on adaptive speed strategy is proposed. This improved algorithm can simulate the bat in the process of search based on adaptive value size and adaptive speed adjustment. His approach can improve the optimization efficiency and accuracy. Experimental results on CEC2013 test benchmarks show that our proposal has better global searchability and a faster convergence speed, and can effectively overcome the problem convergence.

This paper proposes a dual-domain compression sensing (DCS) data collection scheme. The scheme requires only through some nodes for data collection, and it uses the multi-user detection algorithm based on spatial sparse compressive sensing to perform node active state and data detection at the receiver, and then uses the sparsity of the frequency domain for information recovery, thereby further saving the control overhead of the sink node’s downstream sending address frame. Through the comparison of simulation experiments, it is found that the scheme proposed in this paper is better than the previous IDMA multiple access detection schemes in terms of bandwidth while guaranteeing the reconstruction performance of the marine environment monitoring network.

Frequency deviation incurred by sudden changes of frequency introduces harmonics and inter-harmonics in the power system, which influences the accuracy of frequency estimation with the method of discrete Fourier transform (DFT). A two-layered iterative DFT (TLI-DFT) with re-sampling was presented to measure the frequency in non-steady states. A simple frequency estimation method named exponential sampling is amended to calculate the initial sampling frequency in the inner-layered process of the DFT iteration. TLI-DFT can track the frequency in non-steady states that is contaminated by decaying direct current offsets. Mean squared error of measured frequency and rate of change of frequency indicate that the proposed algorithm is valid and more accurate than the traditional one under a transient condition in the power system.

The paradox of huge volume with high sparsity of rating data in collaborative filtering (CF) system motivates the present paper to utilize information underlying sparsity to reduce the dimensionality of data. This difference in user experiences resembles factor underlying widely used term frequency weighting scheme in information retrieval. Hypothesis of Rational Authorities Bias (H-RAB) is proposed, supposing that higher prediction accuracy can be attained to emphasize referential users with higher experiences. Dimension reduction suggests pruning all referential users with less experience than a given maturity threshold. Empirical results from a series of experiments on three major public available CF datasets justify the soundness of both modifications and validity of H-RAB. A few open issues are also proposed for future efforts.

The radar system and communication system in the military and civilian are developing to millimeter-wave band; in order to coordinate the frequency use of communication system and the radar system in this band, it is necessary to analyze the electromagnetic compatibility between them. In this paper, the LFM pulse compression radar in the 35 GHz band is taken as an example and the interference system model is constructed to analyze the electromagnetic compatibility with digital cellular mobile communication system in the same frequency band. The physical-layer simulation is carried out on the millimeter-wave radar and the communication system, and the corresponding interference limits are obtained by combining the path loss, bandwidth, and equipment space distribution of the specific system. Then, an effective link-level interference evaluation model is established, thus obtaining the system transmission power threshold and equipment distribution isolation distance threshold reference value. The research results can provide some technical references for the relevant departments of radio management in China.

A new synthetic aperture radar (SAR) image denoising method based on fast weighted nuclear norm minimization (FWNNM) is proposed. SAR image is firstly modelled by a logarithmic additive model for modelling of the speckle. Then, the non-local similarity is used for image block matching. Next, according to the framework of the low-rank model, randomized singular value decomposition (RSVD) is introduced to replace the singular value decomposition (SVD) in weighted nuclear norm minimization (WNNM) for approximating the low-rank matrix. Finally, the gradient histogram preservation (GHP) method is employed to enhance the texture of the image. Experiments on MSTAR database show that the proposed approach is effective in SAR image denoising and the edge preserving in comparison with some traditional algorithms. Moreover, it is three times faster than WNNM method.

Compressed sensing (CS) theory describes the signal using space transformation to obtain linear observation data selectively, breaking through the limit of the traditional Nyquist theorem. In this paper, we aim at accelerating the current approximate message passing (AMP) and propose an approach named real-valued AMP (RAMP) for faster and better inverse synthetic aperture radar (ISAR) imaging reconstruction. The azimuth dictionary is first processed with real. We then use matrix processing to solve the AMP vector iterative method, by utilizing the relation between the quantification of matrix product and the Kronecker product. The experimental results are presented to demonstrate the validity of this method.

Inverse synthetic aperture radar (ISAR) imaging based on the range-Doppler (RD) imaging algorithm has been proved to be effective at the case of the target moves smoothly. However, for the maneuvering target, due to the Doppler frequency of the received signals, which can be regarded approximately as chirp signals, is time-varying, the conventional RD imaging algorithm is not appropriate anymore. In consideration of the fractional Fourier transform (FRFT) having an advantage in concentrating the energy of the chirp signal, a novel technique of ISAR image formation of the maneuvering target through exploiting the FRFT is proposed. Different from the traditional Wigner–Ville distribution (WVD) and other bilinear time–frequency distributions, the FRFT is a linear operator so that it cannot be affected by the cross-terms for the multicomponent signals. The validity of the introduced method is validated with the results of ISAR imaging for simulated and real data.

In order to make up the shortcomings of the traditional CFAR detection algorithm, a ship target detection algorithm based on information theory and Harris corner detection for SAR images is proposed in this paper. Firstly, the SAR image is pretreated, and next, it is divided into superpixel patches by using the improved SLIC superpixel generation algorithm. Then, the self-information value of the superpixel patches is calculated and the threshold T1 is set to select the candidate superpixel patches. And then, the extended neighborhood weighted information entropy growth rate threshold T2 is set to eliminate false alarm detection results of the candidate superpixel patches. Finally, the Harris corner detection algorithm is used to process the detection result, the number of the corner threshold T3 is set to filter out the false alarm patches, and the final SAR image target detection result is obtained. The effectiveness and superiority of the proposed algorithm are verified by comparing the proposed method with the results of CFAR detection algorithm combining with morphological processing algorithm and information theory combining with morphological processing algorithm on the experimental high-resolution ship SAR images.

In this paper, we propose sense-through-foliage targetdetection approach based on Hidden Markov Models (HMMs). Separate Hidden Markov Models are trained for signals containing target signature and no target (clutter), respectively. Less correlated features are selected as input of Hidden Markov Models for training and testing. Foliage data is collected from three different UWB radar locations, and experimental results show that position 1 data gives the best detection result. All three locations have above 0.8 AUC from the ROC curves.

With the increasing resolution of synthetic aperture radar (SAR), the traditional SAR image target detection methods used for medium-low resolution are not suitable for high-resolution SAR images, which contain detailed information about structure, shape, and weak echoes that are hardly detected in traditional ways. In this paper, we proposed a new method, Superpixel-Random Forest Technique, to detect ships in high-resolution SAR images. The method combines superpixel and random forest algorithms. The superpixel is adopted to divide images into many subregions properly, and the random forest is used for unsupervised clustering these subregions into ships or others. The experimental results show that the algorithm can accurately detect the ship targets.

The detection of vital signs by the ultra-wideband (UWB) radar sensor does not require touching the body. The UWB radar sensor has shorter detection time and less radiation than X-rays. Considering that heartbeat signals and respiratory signals are non-stationary, this paper uses the empirical mode decomposition (EMD) algorithm and the variational mode decomposition (VMD) algorithm to process them. Both respiratory and heartbeat characteristics are obtained with non-contact detection in the paper. The Hilbert transform of vital signs is conducted to reflect the time–frequency information of vital signs. Experimental results show that a highly accurate detection of vital signs can be achieved using the proposed methods.

Intra-pulse modulation recognition is one of the radar reconnaissance key technologies; it is especially a hot point of recent researching under low SNR. This thesis propounds a novel way for radar intra-pulse modulation characteristic recognition based on machine learning means of extreme learning machine (ELM), which is widely applied in the region of pattern recognition. As a novel learning framework, the ELM attracts increasing draws in the regions of large-scale computing, high-velocity signal processing, and artificial intelligence. The aim of the ELM is to break the barriers down between the biological learning mechanism and conventional artificial learning techniques and represent a suite of machine learning methods in which hidden neurons need not to be tuned. This algorithm has a trend to provide perfect generalization performance at staggering learning rate. This article focuses on the high frequency (HF) channel environment and Wavelet transform algorithm with the lower computational complexity. The simulation results imply that the ELM could reap a perfectly satisfactory acceptance performance and therefore supplies a substantial ground structure for dealing with intra-pulse modulation challenges in inadequate channel conditions.

Transmitting beampattern synthesis is of great importance in colocated MIMO radar since the system demands a flexible power transmitting. However, most research focuses on the optimization of transmitting waveform covariance matrix and the design of transmitting waveform with the requirement of arbitrary waveform generators. In this paper, a transmitting beampattern synthesis method based on pulse-to-pulse coding is proposed. The waveform diversity is obtained in Doppler domain since each transmitting signal is modulated to a different Doppler frequency. Then by selecting the additional phase properly, the transmitting beampattern is synthesized to emit power to specific region. This proposed method could reduce the cost of system hardware since all transmitting signals are identical except the phase. Comparing to conventional transmitting beampattern synthesis method, the system complexity and computation complexity are reduced. Simulation results verify the efficiency of our proposed method.

In this paper, an effective approach of ship target recognition is proposed. This method based on the theory of transfer learning aims at using labeled ships with different imaging angles and different resolutions to help identifying unlabeled ships in a fixed angle. Since training ship samples and test ship samples are imaging in different angles, they obey different distributions. However, in traditional machine learning method, training data and test data obey the same distribution. In order to solve this problem, we proposed a method called mapped subspace alignment (MSA) which is different from other domain adaptation methods. While maximizing the difference between different categories, it first uses Isometric Feature Mapping (Isomap) to generate subspace and uses objective functions to spatial alignment and probabilistic adaptation. This paper focuses on the identification of three types of ships which are destroyers, cruisers, and aircraft carriers basing on MSA. The experimental results show that this method is better than several state-of-the-art methods.

Through wall human detection based on ultra-wideband (UWB) radar is a challenging task due to the complex environment. In this case, it is not enough for the research sample that is only with high cost. In this paper, we propose a novel algorithm named support tensor machines (STMs). It avoids the overfitting in pattern recognition. We conduct two groups of experiments on high-dimensional and small-sampling data. The experimental results prove that our method not only achieves the desired results, but also saves plenty of computation time.

To solve the problem of the low recognition rate of the existing methods at low signal-to-noise ratio (SNR), we propose a novel method of radar signal waveform recognition. In this method, we extract the time-frequency images (TFIs) of radar signals through Cohen class time frequency distribution. Then, we introduce convolutional denoising autoencoder (CDAE) to denoise and repairs the TFIs. Finally, we build a convolutional neural network (CNN) to identify the TFIs of radar signals. Simulation experiment shows that the proposed method can identify 12 kinds of radar signal waveforms, and the overall probability of successful recognition (PSR) is 95.4% when the SNR is −7 dB.

This paper proposed a new method to classify soil pH based on long short-term memory (LSTM) via ultra-wideband (UWB) radar echoes. The main contribution of this paper is to provide a solution by incorporating the LSTM into the field experiment related to UWB based on soil pH echoes. Five types of UWB soil echoes with different pH values are collected and investigated using LSTM approach. Finally, the analysis of results shows that LSTM method presents a good classification performance with a short execution time and the data features do not need to be extracted manually. The high accuracy rate also shows that LSTM method is beneficial to the study of other soil parameters.

This paper investigated the compressed sensing (CS) of ultra-wideband (UWB) soil echo signals. When CS is used in the transmission of UWB signals, sampling rate can be effectively reduced and sparse signals can be reconstructed from fewer observations. Therefore, how to apply CS into UWB soil echo signals is of great importance. The proposed approach reveals that UWB signals can be expressed by linear combinations of many atoms from a proper dictionary. In this paper, K-singular value decomposition (KSVD) dictionary and three types of Gaussian pulse dictionaries are designed, and the probability of successful reconstruction can reach 0.95. It is shown that Gaussian first-order derivative dictionary is the most suitable; the root-mean-square error (RMSE) of UWB signals and reconstructing signals is lower than 0.12.

ISAR imaging of dynamic target is very significant in real applications. Plenty of available outcomes have been acquired by the scholars in the near years. Considering the accuracy and the computational complexity, the radar echo can be described as multicomponent QFM signal after envelope alignment with initial phase calibration. Many parametric algorithms in this case have been developed recently. This paper provides a comprehensive summarization of the ISAR imaging approach with QFM signal model in recent years, with the aim to introduce the research progress of it to the researchers and interested readers.

The compressive sensing (CS) technique has been introduced to the field of synthetic aperture radar (SAR) imaging procedure to reduce the amount of measurements. In this letter, a novel algorithm for bistatic SAR imaging based on the CS technique is proposed. The range profile is reconstructed by the Fourier transform, and the azimuth processing is implemented by the CS method consequently. The proposed algorithm can realize the high-quality imaging with limited measurements efficiently for the missing bistatic SAR radar echoes. Results of simulated data demonstrate the validity of the novel approach.

Video synthetic aperture radar (ViSAR) as a new imaging mode has received more attention in recent years. In this mode, ViSAR system needs to persistently form a sequence of images while collecting data. For ViSAR, the frame rate needs to exceed 5 Hz to track maneuvering ground targets. Therefore, ViSAR system generally works in terahertz band. To avoid the 2-D interpolation of the polar format algorithm (PFA), a ViSAR imaging method is proposed in this paper. Based on the small synthetic angle characteristic of THz ViSAR, this method decomposes 2-D interpolation into range and azimuth interpolation, which is implemented by chirp Z-transform (CZT). The effectiveness of this method is validated by the simulation results in 0.3 THz.

For quarter wave vertical antenna in high-frequency band, a ground screen system is indispensable. The ground screen is laid to reduce the effects by earth whose conductive is nonuniform and poor. High-frequency surface wave radar (HFSWR) system demands antenna directivity stricter than the radio station, so the system needs a high-quality ground screen system. This work examines the performance of the ground screen system for high-frequency antenna and finds applicable size and shape for antenna fabrication. Different ground screen sizes, densities and shapes are compared in the simulation for a 9.8 m tall monopole antenna in 7 MHz. The simulation results show that the size and density of the ground screen are important parameters for antenna, and radial ground screen performs better than latticed when the size of the ground screen is large. To find appropriate ground screen parameters for HFSWR, the analysis takes the effect of sea and coast into account. Furthermore, the results may provide guidance for ground screen laying in practical applications.

In view of characteristics of channel parameters changing with time and space of underwater acoustic channel, a rateless transmission scheme for adaptive transmission over underwater acoustic channels is studied in this paper. The scheme utilizes the characteristics of LT coding without fixed coding rate, and the communication efficiency is further improved through single-carrier frequency-domain equalization. Simulation results show that the scheme we proposed can effectively reduce the computational complexity compared to the rateless coding system under classical time-domain equalization. With achieving the same reliable information transmission in the same underwater acoustic channel, the redundancy required by the proposed scheme is about 1/3 of the classical retransmission method, which improves the efficiency of information transmission.

The variation of rotary joint is one of the primary reasons that result in dynamic deviation of ZDR and ΦDP as polarimetric parameters, and external instruments are needed to detect and calibrate to ensure the amplitude and phase consistency for dual-polarization weather radar. This paper analyzes the impact of rotary joint of dual-polarization weather radar on ZDR and ΦDP, and it proposes a detection and calibration method using external instrument based on baseline curve of deviation. Then this method is used to test and calibrate a S-band dual-polarization weather radar produced by Beijing Metstar Radar Co., Ltd., and the results are analyzed. It is shown that ZDR and ΦDP deviation introduced by rotary joint can satisfy requirement of relevant technical specifications, and this method can reduce deviation of amplitude and phase through calibration to enhance the reliability of radar observation data effectively.

Pulse ultra-wideband technology uses narrow pulse as the information carrier, occupies a very wide frequency bandwidth, and has many characteristics such as high transmission rate, large capacity and strong multipath resistance. Chaotic signals are stochastic and sensitive to initial values, so they are widely used in secure communication systems. In this paper, the chaotic sequence of quadratic chaotic map is generated by matrix method. The new digital chaotic sequence after binary quantization is applied to the direct sequence spread spectrum UWB system. It reduces the influence of quantization on the randomness and ameliorates the dispersion of UWB signals power spectrum, making the interference with other systems lower. The simulation shows that the chaotic direct sequence spread spectrum UWB system based on chaotic matrix quadratic mapping has better system coexistence.

Due to the limited energy of nodes and the harsh working environment in underwater sensor networks, designing energy-efficient routing algorithms to achieve data acquisition is particularly important. Using the correlation of original signal in underwater sensor networks, in this paper, an uneven-layered, multi-hop routing based on distributed compressed sensing (DCS-ULM) is proposed to achieve data collection. The simulation results show that DCS-ULM can effectively prolong the lifetime of networks while ensuring the reconstruction accuracy of original data.

The signal can be charactered by both eigenvalues and eigenvectors of covariance matrix. However, the existing detection methods only exploit the eigenvalue or eigenvector. In this paper, we utilize both eigenvalues and eigenvectors of the sampled covariance matrix to perform spectrum sensing for improving the detection performance. The features of eigenvalues and eigenvectors are considered integratedly, and the relationship between the false-alarm probability and the decision threshold is offered. To testify this method, some simulations are carried out. The results demonstrate that the method shows some advantages in the detection performance over the conventional method only adapting eigenvalues or eigenvectors.

The mean-shift (MS) tracking is fast, is easy to implement, and performs well in many conditions especially for object with rotation and deformation. But the existing MS-like algorithms always have inferior performance for two reasons: the loss of pixel’s neighborhood information and lack of template update and scale estimation. We present a new adaptive scale MS algorithm with gradient histogram to settle those problems. The gradient histogram is constructed by gradient features concatenated with color features which are quantized into the 16 × 16 × 16 × 16 bins. To deal with scale change, a scale robust algorithm is adopted which is called background ratio weighting (BRW) algorithm. In order to cope with appearance variation, when the Bhattacharyya coefficient is greater than a threshold the object template is updated and the threshold is set to avoid incorrect updates. The proposed tracker is compared with lots of tracking algorithms, and the experimental results show its effectiveness in both distance precision and overlap precision.

Due to the poor robustness and high complexity of the concentrated dictionary learning (CDL) algorithm, this paper addresses these issues using denoising deep extreme learning machine based on autoencoder (DDELM-AE) and approximate k singular value decomposition (AK-SVD). Different from the CDL algorithm, on input, DDELM-AE is added for enhancing denoising ability and AK-SVD replaces K-SVD for improving running speed. Additionally, experimental results show that the improved algorithm is more efficient than the original CDL algorithm in terms of running time, denoising ability, and stability.

Recently, researchers around the world have been striving to develop human–computer interaction systems. Especially, neither special devices nor vision-based activity monitoring in home environment has become increasingly important,and has had the potential to support a broad array of applications. This paper presents a novel human dynamic gesture recognition system using Wi-Fi signals. Our system leverages wireless signals to enable activity identification at home. In this paper, we present a novel Wi-Fi-based body gestures recognition model by leveraging the fluctuation trends in the channel of Wi-Fi signals caused by human motions. We extract these effects by analyzing the long training symbols in communication system. USRP-N210s are leveraged to set up our test platform, and 802.11a protocol is adopted to implement body gestures recognition system. Besides, we design a novel and agile segmentation algorithm to reveal the specific pattern and detect the duration of the body motions. Considering the superiority of feature extraction, convolutional neutral networks (CNN) is adopted to extract gesture features, and extreme learning machine (ELM) is selected as classifier. This system is implemented and tested in ordinary home scenario. The result shows that our system can differentiate gestures with high accuracy.

In this paper, we mainly evaluate the influence of local features extracted by convolutional neural networks for person re-identification. Considering the variant body parts with different structural information, we divide the holistic person images into several parts and extract their features. Two kinds of aggregation methods are used to aggregate local features. Experiments on the challenging person re-identification database, Market-1501 database, show that the max aggregation is more effective for extracting the discriminative local features than the sum aggregation.

In this paper, we propose a new method for modulation recognition of received digital signals using bispectrum and AlexNet. The bispectrum analysis is used to generate the feature images, AlexNet, as a widely used deep neural network (DNN), is used as the classifier. It is able to classify six common digital communication signals, including 2ASK, 4ASK, 2FSK, 4FSK, 2PSK and 4PSK. Compared to the traditional decision-theoretic methods, the proposed method needs no prior information for the received signals. The numerical results indicate that this method is more robust and effective than the classical decision theory and its improved algorithm, particularly when the signal-to-noise ratio (SNR) is low. It is shown that the success rate of 90% can be achieved when the SNR is greater than or equal to 3 dB.

Image-to-image translation is a hot field in the machine learning with the emergency of the generative adversarial networks. Most of the latest models easily lead to changes in the overall image and overfitting when they are used to local feature translation. To address these limitations, this article adds a suppressor and proposes a double adversarial CycleGAN. The suppressor is added to suppress the change of images, and the suppressor and generator form a new adversarial relationship. We hope it will achieve Nash equilibrium that is the change of image focus on the local feature. Finally, a contrast experiment was conducted. In the case of image local feature transfer, the change of image is focused on the local features and the overfitting phenomenon can be well resolved.

Ground-based cloud classification plays a vital important role in meteorological research. However, the existing methods perform well confined to one weather station. In this paper, we present a detailed introduction of two representative embedding features for ground-based cloud classification in various weather stations. The features are learned from the metric learning and the convolutional neural network (CNN), respectively. The two kinds of features are evaluated on two weather stations.

In this paper, we proposed a novel DCT-based watermarking scheme for grayscale images, which utilizes the connection between discrete cosine transform (DCT) and the Histogram of Oriented Gradient (HOG) feature extraction operation. We embed messages into the low-frequency band, and correspondingly, an effective coefficients pair selection scheme is constructed. The proposed scheme not only maintains the superiority of compression robustness but also keeps good visual quality for the watermarked image. Moreover, the proposed method is insensitive to HOG feature extraction, which makes the watermarked image more suitable for further objection detection and recognition scenarios.

To solve the problem of reconnaissance and processing of broadband satellite communication signals, a kind of satellite communication signals BPSK/QPSK modulation pattern recognition method was put forward in this paper. This method deals with the satellite descending signal with BPSK/QPSK modulation in the under-sampling condition. Because the corrected spectrum of BPSK signal contains obvious crest, while QPSK signal does not contain this feature. The difference of the waveform characteristics is used to complete modulation pattern recognition. The simulation results show that this method can identify BPSK/QPSK modulation signals when SNR is greater than 1 dB. When the sampling points are reduced, the satellite communication signal under-sampling modulation pattern recognition method can still maintain good recognition performance.

Given the revival of neural networks and its recent impact in other disciplines and record-breaking performances in a variety of applications, in this paper, we employed a deep sequential model for polyps detection from the vocal data. Previous research of acoustic signal recognition (ASR) has focused on hand-crafted machine learning fashion, such as Mel-frequency cepstral coefficients with hidden Markov model and Gaussian mixture model. The deep model demonstrates its flexibility and potential to outperform the traditional methods, and we expand its scope on medical symptom identification. The mapping between the raw vocal signal and the symptom recognition is established, and we show that we can achieve a good recognition accuracy, which may appear to clinical diagnosis in the near future.

Audio tagging aims to predict one or several labels in an audio clip. Many previous works use weakly labelled data (WLD) for audio tagging, where only presence or absence of sound events is known, but the order of sound events is unknown. To use the order information of sound events, we propose sequentially labelled data (SLD), where both the presence or absence and the order information of sound events are known. To utilize SLD in audio tagging, we propose a convolutional recurrent neural network followed by a connectionist temporal classification (CRNN-CTC) objective function to map from an audio clip spectrogram to SLD. Experiments show that CRNN-CTC obtains an area under curve (AUC) score of 0.986 in audio tagging, outperforming the baseline CRNN of 0.908 and 0.815 with max pooling and average pooling, respectively. In addition, we show CRNN-CTC has the ability to predict the order of sound events in an audio clip.

For the problem that Adaptive Boosting (AdaBoost) face detection algorithm is slowly implemented on the embedded platform by software, this paper adopts the method of the full hardware acceleration. The intellectual property (IP) core of AdaBoost algorithm is designed by Vivado high-level synthesis (HLS), which may reduce the development difficulty and shorten the development cycle. The design adopts the serial–parallel structure to accelerate face detection and uses several methods of optimizing hardware resource. The face detection algorithm is implemented on the Zedboard platform and achieves the purpose of real-time detection.

In the paper, a novel rolling bearing fault diagnostic method was proposed to fulfill the requirements for effective assessment of different fault types and severities with real-time computational performance. Firstly, multi-dimensional feature extraction is discussed. And secondly, a gray relation algorithm was used to acquire basic belief assignments. Finally, the basic belief assignments were fused through Yager algorithm. The related experimental study has illustrated the proposed method can effectively and efficiently recognize various fault types and severities.

In order to address the impact of large-angle posture changes on face recognition performance, we propose a contour symmetric constraint-generative adversarial network (CSC-GAN) for the multi-pose face recognition. The method employs the convolutional network as the generator for face pose recovery, which introduces the global information of the constrained pose recovery of positive face contour histogram. Meanwhile, the original positive face is used as the discriminator, and the symmetric loss function is added to optimize the learning ability of the network. The positive face with gesture recovery is obtained by striking the balance between training of the generator and discriminator. Then we employed the nearest neighbor classifier to identify. The experimental results show that CSC-GAN obtained good posture reconstruction texture information on the multi-pose face reconstruction. Compared with the traditional deep learning method and 3D method, it also achieves higher recognition rate.

The purpose of this research is to increase the target recognition rate by means of neural networks and feature fusion. We analyze the performance of different recognition methods (Bayesian classifier, support vector machine (SVM), and neural networks) based on high-resolution range profile (HRRP). The result shows the superiority of neural networks to Bayesian classifier and SVM in classification. We apply multi-source feature fusion to target recognition based on neural networks. The results show that, in certain cases, the target recognition ratio using fusion feature is higher than that of HRRP only.

For the high dimension of fingerprint feature set in the process of specific emitter identification (SEI), feature selection method is utilized to reduce the feature dimension and improve individual recognition rate. This paper adopted the filter feature selection in four ways: MIFS, mRMR, CMIM, and JMIM fingerprint feature set of high-dimensional feature selection and combined with PCA dimensionality reduction algorithm to minimize the feature dimension. The simulation results show that feature selection is feasible in individual recognition of the radiation source and can be effectively combined with dimension reduction algorithm.

Extracting applicable features from continuous gesture is uneasy since it shows up as a nonlinear dynamic system with a spatial–temporal pattern. This paper introduces a continuous gesture recognition framework that analyzes, models, and classifies the nonlinear dynamics of gestures based on chaotic theory. In this system, the trajectories of finger joints are captured as the discrete observations of nonlinear dynamic system, which defines the feature matrix of gestures by reconstructing a phase space through employing a delay-embedding scheme, the properties of the reconstructed phase space are captured in terms of dynamic and metric invariants that include Lyapunov exponent, correlation integral, and fractal dimension. Finally, we extract a feature matrix for training several classifiers with relatively few samples and get best accuracy of around 96.6% to prove our assumption that the nonlinear dynamics of continuous gesture can be approximated by a particular type of dynamical system for classification.

In order to improve the measuring precision and stability of transit-time ultrasonic flowmeter as well as the locating accuracy of datum point for ultrasonic received signal, a feature wave recognition-based signal processing method is proposed in this study, which derives from analyzing the cause of errors in conventional threshold approach. By introducing a phase-shifted pulse into the ultrasonic excitation one, a feature wave with different period and phase is consequently produced in the ultrasonic received signal and recognized using a high-precision TDC chip according to the period of the received signal at first. Then the datum point of the received signal is accurately located with regard to the relationship between the position of feature wave and the initial position of the received signal so that the transit time of ultrasonic signal is finally measured. The following experiments focusing on a real-world problem demonstrate that the proposed method can effectively reduce the measurement errors caused by the amplitude change of the received signal. Such an approach is greatly beneficial for improving the precision of measurement along with the stability of the ultrasonic flowmeter.

In order to solve the problems of difficult wiring, poor flexibility, and high cost in aquaculture water quality monitoring, an unmanned water quality monitoring cruise ship with water quality monitoring device was built. The ship navigated automatically on the surface of the water according to the set course and collected water quality data during the voyage, which saved a lot of resources in this innovative way. There are two modes of operation for cruise ships: manual mode and autopilot mode. In the manual mode, the user can realize the manual operation of the ship through the remote controller and can set the autopilot path of the cruise ship. In the autopilot mode, the cruise ship moves automatically according to the preset path.

The traditional target activity regular pattern extraction methods replay previous target tracks, activities of the specified target are manually analyzed by checking all the tracks on map. This paper adopts big data mining technology to solve the problem of automatically extracting target classic tracks and converts the original pure manual map analysis into system automatic track extraction. This method greatly reduces the operation intervention of classic track extraction, which can reduce the 3–4 manual days to 3–4 h.

Target detection in the background of sea clutter is an important part of sea surface radar signal processing. The traditional detection of weak targets in sea clutter is based on the statistical characteristics of sea clutter, which does not reflect the intrinsic dynamics of sea clutter. Therefore, the detection results are not ideal. Based on the chaotic characteristics of sea clutter, this dissertation reconstructs the space structure of the sea clutter and proposes an improved particle swarm optimization (PSO) algorithm based on adaptive time-varying weights and local search operators. This method was applied to the optimization learning of the parameters of the radial basis function (RBF) neural network kernel function. The method was validated by using McIX University in Canada to measure the sea clutter data with the target in the Dartmouth area using IPIX radar. The results showed that the PSO-RBF algorithm in the background of chaotic sea clutter has good predictability. Compared with the general radial basis neural network, the improved algorithm not only has fast convergence speed but also has high error accuracy.

In order to improve the reliability and security of biometric-based identity authentication system and reduce the risk of unauthorized access caused by forgery feature attacks, this paper proposes a method for identifying the identity of visitors. The method is based on D-S evidence theory. The palm print and palm vein are used as authentication features. Firstly, the same collection device is used to collect palm print and palm vein images under different wavelengths of light source and extract the HOG features of the image; then, use the one-vs-one multi-classification method of SVM to classify different individuals, and finally, using the D-S fusion strategy at the decision-making level to improve the security and accuracy of the identity authentication system. Through many experiments, the recognition rate of decision-making layer fusion is above 98%, which confirms the effectiveness of the proposed method.

Ice and snow have a detrimental effect on the transmission and distribution lines. In the calculation of the ice thickness of the ice-covered transmission lines by the image method, the center position of the target transmission lines is crucial for the detection result. The detection of ice-covered transmission lines is affected by external factors, resulting in images that are mostly inclined ice-covered images. Therefore, it is necessary first to straighten the wire and then by fitting the image rotation technique.

The relationship of decision rule of sensor for each other is relevant to data fusion, so different topological networks of sensors usually results in different performances. This paper considers the sequential network fusion with two sensors in some detail and compares its performance with that of single detection and fusion. In this paper, the detection model is specified for binary hypotheses testing problem. In particular, this paper supposes that Bayesian risk cost of different decisions and the prior probability distribution of two hypotheses are known. Finally, this paper simulates the probabilities of error and Bayesian risk by these fusion rules with corresponding to different values of prior probabilities of two hypotheses by these fusion methods. And compared to single detection and fusion, the performance of sequential detection and fusion is better.

Generally, particle filter is used in the single snapshot situation and the initial state is assumed to be known. To make the measurement interval be small enough, we construct a multiple measurement vectors model for DOA tracking since it usually outperforms the single measurement vector model. And we take the initial state into consideration. The initial tracking error of the particle filter becomes very large when the initial state is unknown. Thus, we modify the initial state according to the likelihood of the generated random samples. The method is numerically evaluated using a uniform linear array in simulations. The results show that the proposed algorithm has higher tracking accuracy.

Localization is an important requirement in today’s world, and numerous modern applications require location tracking. An indoor localization of a mobile node using the range-free fingerprinting technique in WLAN environment is presented. The work focuses on improvement in the accuracy of localization using some additional parameters in the fingerprint, along with the conventional received signal strength (RSS). ToA has been used to enrich the fingerprint data for more unique fingerprints. The impact of AP placement on localization accuracy is also addressed. In this paper, a technique is proposed that is not complex to implement using existing infrastructure and is also easy to understand. Significant improvement has been achieved from about 20 to 40% in different scenarios including line of sight and non-light of sight scenarios, small and large areas.

This paper proposes a new kernel function in state coherence transform to perform multiple time difference of arrival estimation in order to increase the resolution of location in frequency-domain blind source separation. The state coherence transform associated with each source generalizes the GCC for multiple sources and generates envelopes with clear peaks corresponding to the maximum-likelihood TDOAs. However, the weight allocation of the kernel function is unreasonable for small spacing microphones. We propose an improved kernel function to enhance the resolution of small values, which means that a larger weight allocated to smaller values. Experimental results show that the proposed approach allows to separate four speakers, using very short utterances, in highly reverberant environment even with small-spaced microphones of 2 cm.

In order to solve the problem of parking lot difficulty, communication security and low efficiency of garage, with the garage using LED lighting, designed an underground garage navigation system based on visible light communication, which uses the lighting system in the garage to realize the real-time monitoring of the parking space and the navigation of the vehicle. This paper designed the navigation system of underground garage based on the principle of visible light communication in the garage with LED lighting. Result shows that the system achieves signal transmission in the range of 4 m, which can meet the needs of vehicle navigation.

Recently, GNSS has been applied in various domains deeply and widely. In some of applications such as carbon canyon, GNSS signals degrade severely. The conventional receivers have no ability to deal with such weak signals. The sensitivity performance has already been one of the most important features in modern receivers. Consider the advantages of easy implement and high efficiency, we choose it with coherent integration and differential coherent integration to acquire signals which of power is −145 dBm.

When single satellite navigation system is used for positioning, there exist the following problems: accuracy of positioning is low and reliability of positioning is also low. This paper investigates the combined positioning algorithm of BDS and GPS for static and dynamic observation point. For static observation point positioning, the weight coefficient integrated positioning algorithm of BDS and GPS is proposed. For dynamic observation point positioning, the Kalman filter can achieve smoothing of the movement trajectory which is the once combined observation point. The experimental results show that the combination of BDS and GPS is more accurate and reliable than any single system for static observation point positioning. At the same time, for the consideration of the weight coefficient, the system has good adjustability and practicality, and the Kalman filter can better modify the dynamic combined observation point.

In the paper, an indoor positioning scheme combining pedestrian dead reckoning (PDR) and magnetic strength matching (MSM) is proposed. PDR is conducted by sensing acceleration and angular speed through the 3-axis accelerometer in iphone7 and a fabricated high-precision rotational gyroscope. Low bias stability (0.5°/h) of the gyroscope contributes to a small accumulative error in heading angle estimation. Through data analysis to outputs of the accelerometer and the gyroscope, human motion, such as walking a step, walking upstairs or downstairs, turning left or right, is recognized and walking path is reckoned with motion information. Magnetic strength is measured by the magnetometer in iphone7 and MSM positioning result is used to reduce error of reckoned heading angle. The error rate of downstairs/upstairs step count is low and after heading angle correction by MSM, a satisfactory indoor positioning result is obtained.

This paper designs the SPI interface module in the baseband chip of the dual mode satellite navigation receiver based on the AMBA bus, and this paper is based on the first edition of the baseband chip accepted by the Science and Industry Corp. This design provides the IP core of SPI for the SoC using the LEON series processor and uses the TMR on register transfer level and evaluates the final results after the reinforcement. The results of verification and evaluation show that the SPI master designed by this paper can communicate normally based on the AMBA bus, and the SPI master after reinforcement can resist SEU to a certain extent and improve its stability in the space radiation environment.

The performance of indoor localization algorithm is limited by non-line-of-sight (NLOS) error, a positioning system includes Bluetooth module, Bluetooth gateway and cloud monitoring center based on particle filter is presented to enhance positioning accuracy. Our experimental results indicate that the proposed localization scheme leads to higher localization accuracy and lower power consumption.

This paper focuses on the problem of sensor selection in time-difference-of-arrival (TDOA) localization scenario with correlated measurement noise. The challenge lies in how to select the reference sensor and ordinary sensors simultaneously when the TDOA measurement noises are correlated. Specifically, the optimal sensor subset is found by introducing two independent Boolean selection vectors and formulating a nonconvex optimization problem, which motivates to minimize the localization error in the presence of correlated noise and energy constraints. Upon transforming the original nonconvex problem to the semidefinite program (SDP), the randomization method is leveraged to tackle the problem, and thereby proposing the novel algorithm for sensor selection. Simulations are included to validate the performance of proposed algorithm by comparing with the exhaustive search method.

In the TDOA location system,the constellation drift can affect the location precision. Aiming at this problem, based on the analysis of the three satellites constellation drifting in a period of time, we proposed the location precision results of constellation drift influence in TDOA location system. Considering the constraint of location accuracy and satellite resources, a feasible recommendation for location maintenance of three satellites system is put forward, which provides the theoretical support for the location maintenance strategy of three satellites TDOA location.

In order to solve the problem of the influence of signal strength fluctuation on indoor positioning accuracy, an improved indoor positioning algorithm based on WiFi signal strength is proposed in the paper. Based on the K-nearest neighbor location algorithm, the weight of the signal strength is further increased, the characteristics of the received signal and the fingerprint database are optimized, the interference of the weak signal on the positioning is reduced, and the accurate indoor positioning is achieved. The calculation in this work suggests that the positioning error can be reduced on the original basis, the accuracy of the algorithm is improved. On the basis of the original algorithm, the error range is narrowed and the positioning accuracy is improved. The average error of the improved algorithm is controlled at about 1.87 m.

Pedestrian retrieval is an important technique of searching for a specific pedestrian from a large gallery. In this paper, we introduce three types of distance metrics for pedestrian retrieval, including learning-free distance metric methods, metric learning methods, and convolution neural network (CNN) methods, and evaluate the performance of different distance metrics using the Market-1501 database. The experiment shows that the CNN methods achieve the best results.

In order to improve the accuracy and tracking performance of the indoor positioning system based on visible light communication (VLC), an indoor positioning and tracking method is proposed in this paper. This method utilizes time difference of arrival (TDOA) solved by nonlinear least squares (NLLS) method to realize indoor positioning and uses Kalman filter to obtain the tracking capability. The performance of the proposed positioning method is evaluated in the room measuring 5 m × 5 m × 3 m. The simulation results show that the average location errors by adopting the NLLS method can reach to 2.99 cm and the accuracy of positioning can be promoted to 1.33 cm by using Kalman filter, the positioning accuracy increased by 55.52%.

Massive MIMO is one of the enabling technologies to cope with exponential data growth. It is very crucial for downlink precoding to accurately estimate direction-of-arrival. A lot of work has been done for 2D DOA estimation based on uniform rectangular array. However, in cases that snapshots are severely limited, such as extremely complex communication environment, conventional 2D DOA estimation method cannot work properly. Iterative adaptive approach (IAA) is one of the sparse algorithms which can handle heavy snapshot limitations. In this paper, we propose a novel 2D DOA estimation algorithm based on IAA for massive MIMO systems. Unlike conventional methods, an estimator in this algorithm is updated by previous iteration instead of the snapshots. The iteration ends until convergence. Simulation results demonstrate that the proposed algorithm is superior to conventional methods in low snapshot cases.

Mobile communication is coming to the fifth generation (5G) networks. In the age of 5G, the three-dimensional (3D) beamforming is one of the highlighted technologies. 3D beamforming increases vertical dimension in terms of space domain, but also has brought a challenge to beamforming, especially the problem of direction-of-arrival (DOA) estimation. In this paper, a DOA estimation method based on multiple signal classification (MUSIC) of uniform circular array (UCA) with mutual coupling compensation is presented for the future 5G networks. 2D MUSIC does well in estimating DOA without an accurate number of signals. The MATLAB software is employed in order to conduct the modeling and 2D MUSIC algorithm simulations.

With the extensive applications of unmanned aerial vehicle (UAV), typical algorithm for path planning is usually restricted for its low efficiency and easy failure, especially for the complex obstacle environments. Therefore, in this paper, a new UAV path planning algorithm is proposed based on ant colony optimization (ACO) for such complex obstacle environment. In particular, the proposed algorithm optimizes the distribution of pheromones and modifies the transfer probability by considering the regional security factors. As a result, it can increase search speed and avoid local optimum and deadlock. Simulation results verify the feasibility and effectiveness of the proposed method.

In this paper, we propose a framework for abnormal event detection and analysis in the field of visual surveillance based on the state-of-the-art deep learning techniques. We train a pair of conditional generative adversarial networks (cGANs) using the normal behavior samples, where one cGAN takes video frames as inputs and generates the corresponding optical flow features. While on the other hand, the other cGANs take optical flow features as inputs and generate the corresponding video frames. By analyzing the differences between the generated frames/optical flow features and the realistic samples, abnormal events can be detected and localized effectively. Moreover, for suspected regions, we adopt the faster RCNN to analyze the abnormal events. Experimental results demonstrate that the proposed framework can detect the abnormal events accurately and efficiently.

The multi-mode positioning of Global Navigation Satellite System (GNSS) could improve the positioning accuracy compared with the traditional single-mode positioning, as the number of observed satellites is increased and the geometry distribution of visual satellites is improved. In this paper, pseudorange fusion algorithm is proposed to combine the pseudorange observations of both Global Positioning System (GPS) and BeiDou Navigation System (BDS) to obtain the position equation for dual-mode positioning. Then the weighted least square method is used to solve this position equation. Besides that, the proposed pseudorange fusion algorithm is implemented in a GPS/BDS software receiver. According to positioning result comparison of single-mode and dual-mode positioning, it is concluded that the dilution of precision (DOP) of the dual-mode positioning is smaller and the positioning accuracy is more precision.

Wireless localization has become a hot issue in Internet of things, but the none-line-of-sight (NLOS) propagation will degrade the performance of traditional localization algorithms. Therefore, this paper proposed an improved range scaling algorithm (RSA) in the wireless sensor networks, where we use a two-step improvement to enhance the constrained optimization model. Simulations demonstrate that the proposed algorithm outperforms the compared algorithms, and effectively suppress the localization error caused by the none-line-of-sight propagation.

With the popularity of smart devices, applications based on location services have been widely used, and wireless positioning technology can provide accurate positioning information. However, due to the effect of non-line-of-sight (NLOS) errors, the performance of the system can drop significantly. Accordingly, this paper introduces the theory of quadratic programming optimization based on the research of the time difference of arrival (TDOA) theory and proposes an optimization algorithm that can effectively suppress the influence of NLOS error. Simulation results show that compared with other common wireless location algorithms, the proposed algorithm has more reliable positioning accuracy under different environment models and has better system stability.

This paper studies an indoor positioning method combining wireless sensor network (WSN) and inertial navigation system (INS). Because the positioning error of INS increases with time, the long-term positioning accuracy is poor, so the combination uses the wireless sensor network to measure the distance between unknown node and base station by TDOA method. The dead-reckoning data of inertial measurement unit (IMU) and the distance information of TDOA method are transmitted to the processing terminal, and then the particle filter algorithm is used to smooth the data to obtain the position estimation. The cumulative positioning error of INS is corrected, and the non-line-of-sight (NLOS) error in TDOA positioning method is reduced. The experimental results show that compared with the single TDOA localization method, the accuracy of the combined positioning method is higher.

In the array signal processing, the direction of arrival (DOA) is a very important parameter. It can be used to estimate the spatial parameter or source location of signals, so it has been deeply studied by scholars and has been widely applied in radar detection, underwater operation, and mobile communication. In this paper, an array signal DOA estimation and display system are proposed. It adopts the fast beamforming algorithm, which is an improved algorithm based on the MVDR algorithm. By comparing the output power of different positions, the angle of the arrival of the signal to the array can be obtained. The system can manually set the basic parameters, such as the number of sensors, the frequency of sampling, the parameter of the filter, the transmission speed of the sound wave, and so on. Then, the waveform characteristics of the time and frequency domain of the original acquisition signal and the filtered signal are displayed intuitively, and the real-time waveform of the signal is displayed at the same time. Through theoretical analysis and simulation experiments, we can get that the related algorithm in this system is not only suitable for microphone array, but also can meet the application requirements of radar monitoring, underwater operation, and other fields.

In the process of direction-of-arrival (DOA) estimation, the difference co-array of sparse arrays can achieve high degrees of freedom, which can be utilized to detect more signal sources than physical sensors based on spatial smoothing (SS) algorithm. In this paper, we present a method for DOA estimation using sparse signal recovery through compressive sensing (CS) approach in the presence of mutual coupling. Compared with SS algorithm, CS approach achieves a lower estimation error. Additionally, simulation results show that the estimation error of CS approach increases with the increase of mutual coupling. Also, it increases with the increase of the grid interval of the entire DOA space.

The vast market of location-based services (LBSs) has brought opportunities for the rapid development of indoor positioning technology. In current indoor venues, by considering the fact that the wireless local area network (WLAN) infrastructure is widely deployed, the indoor WLAN localization method has become the focus of study. Nowadays, the WLAN module is used widely in a large number of advanced mobile devices, and meanwhile there are a variety of WLAN mobile access points (APs) in indoor environment. In this circumstance, due to the uncertainty of the state of mobile APs, the associated received signal strength (RSS) data are usually lowly dependent on the locations, which will consequently result in the decrease in localization accuracy. To solve this problem, a new method of mobile AP detection based on the density-based spatial clustering of applications with noise (DBSCAN) is proposed. This method aims to identify mobile APs in target area so as to eliminate the adverse impact of mobile APs on localization accuracy.

Applications on location-based services (LBSs) have driven the increasingly demand for indoor localization technology. Motivated by the widely deployed wireless local area network (WLAN) infrastructure and the corresponding easily accessible WLAN received signal strength (RSS) data, the Wi-Fi signal-based localization has become one of the superior positioning techniques in GPS-denied scenes. Meanwhile, the error bound estimation for the Wi-Fi localization has been attracting much attention due to its significant guidance meaning in practice. In this survey, the error bound estimation approaches for different categories of Wi-Fi localization approaches are overviewed and compared, including the error bound estimation with temporal and spatial signal features, and that with the RSS characteristics. Regarding the temporal and spatial signal feature-based Wi-Fi localization, we present how to utilize the time of arrival (TOA), the time difference of arrival (TDOA) as well as the arrival of angle (AOA) to analyze the error bound of localization systems. Regarding the received signal strength (RSS) characteristic-based Wi-Fi localization, we clarify the error bound estimation approaches for both the wireless signal propagation-based and location fingerprinting-based localization schemes. In addition, some future directions with respect to the error bound estimation for Wi-Fi localization are also discussed.

With the dramatic development of location-based service (LBS), indoor localization techniques have been widely used in recent years. Among them, the indoor wireless local area network (WLAN) localization technique is recognized as one of the most favored solutions due to its low maintenance overhead and high localization accuracy. In this paper, we propose a new received signal strength (RSS)-based indoor localization approach using augmented manifold alignment. First of all, we construct the objective function in manifold space for indoor localization. Second, the optimal transform matrix is used to transform the coordinates of reference points (RPs) and the corresponding RSS vectors into manifold space. Finally, we locate the target at the RP with the transformed coordinates nearest to the transformation of the newly collected RSS vector in manifold space. The experimental results demonstrate that the proposed approach is able to achieve satisfactory localization accuracy with low overhead.

The traditional outdoor integrated positioning and navigation system is normally suffered by the disadvantages of accumulative error and high power consumption. To solve this problem, we propose a new trajectory reckoning method which use the BeiDou system (BDS) to conduct the attitude measuring and point positioning with respect to the target. In concrete terms, the target location is estimated by solving the pseudo-range observation equation, while the attitude angle is obtained from the dual-difference pseudo-range and carrier phase observation equations. Then, the trajectory of the target is constructed based on the estimated location and associated attitude angle. Finally, the extensive experimental results demonstrate the effectiveness of the proposed trajectory reckoning method with the BDS attitude measuring and point positioning.

Indoor WLAN intrusion detection technique for the anonymous target has been widely applied in many fields such as the smart home management, security monitoring, counterterrorism, and disaster relief. However, the existing indoor WLAN intrusion detection systems usually require constructing a passive radio map involving a lot of manpower and time cost, which is a significant barrier of the deployment of WLAN intrusion detection systems. In this paper, we propose to use the adaptive-depth ray tree model to automatically construct an adaptive passive radio map for indoor WLAN intrusion detection. In concrete terms, the quasi-3D ray-tracing model is enhanced by using the genetic algorithm to predict the received signal strength (RSS) propagation feature under the indoor silence and intrusion scenarios, which improves the computational efficiency while preserving the accuracy of passive radio map. Then, the RSS mean, variance, maximum, minimum, range, and median are allied to increase the robustness of passive radio map. Finally, we conduct empirical evaluations on the real-world data to validate the high intrusion detection rate and low database construction cost of the proposed method.

In order to improve the performance of the existing iris location algorithm, a transform algorithm based on gray projection and Hough is proposed. The algorithm uses the grayscale transformation of the binary image to obtain a graph of the gray projection. At the same time, according to the value of the peak or trough in the graph, the maximum radius of the circle is obtained. The result of experiment shows that: The algorithm can get the parameters needed in Hough transform, which greatly improves the speed and accuracy of iris positioning.

In this paper, we propose a novel correlation filter framework constrained by dual filters. The Minimum Output Sum of Squared Error (MOSSE) filter is the unbiased estimate of the filter which easily to cause overfitting. The trained filter by linear ridge regression is the biased estimate of the filter which can deal with the overfitting. We combine the advantages of the two filters to constrain the trained filter which optimizes our model. To deal with background clutter, clipping background patches around the target position up, down, left, and right, we add the cropped background patches to the learning filter. To overcome the challenge of occlusion, we introduce a novel criterion, Average Peak-to-Correlation Energy (APCE). Extensive experiments on the CVPR 2013 Benchmark well demonstrate that our tracker can effectively solve the background clutter and occlusion. Both quantitative analysis and qualitative analysis show that our tracker outperforms some state-of-the-art trackers.

Localization is an important requirement for wireless sensor networks (WSNs), but the inclusion of GPS receivers in sensor network nodes is often too expensive. Therefore, many solutions focus on static networks and do not consider mobility. In this paper, we analyze the Monte Carlo location (MCL) algorithm and propose an improved method—grid-based MCL. It applies the mobility of nodes to reduce the sampling area and to build an internal grid to predict the behavior of nodes. We investigate the properties of our technology and analyze its performance. The simulation and analysis show that the proposed grid-based MCL not only reduces localization error, but also improves the sampling efficiency.

In indoor localization scenarios, a sheer coordinate with respect to a basis is insufficient to indicate the users’ situation due to a lack of information about landmarks distributed in the environments. To extract landmarks’ information manually, however, is inefficient and thus vulnerable to changes of the environments. Simultaneous localization and mapping can solve the localization and landmarks’ information extracting problems. This paper presents WalkSLAM, a SLAM solution that estimates both the path taken by the user and the locations of Wi-fi devices in the indoor space, using a smartphone. This solution extends the previous work by introducing human walking patterns into the specific SLAM problem. Experiments demonstrate that the improvement consists of increased efficiency of the particle filter, and hence, of the overall algorithm, and a better estimation of the user’s location and path.

In this paper, the time-frequency spatial smoothing MUSIC algorithm (TF-SSMUSIC) for DOA estimation based on co-prime array is proposed. The spatial smoothing MUSIC (SSMUSIC) is a typical DOA estimation algorithm based on co-prime array. TF-SSMUSIC replaces SSMUSIC’s data covariance matrix with a time-frequency distribution matrix, which leads to a better DOA estimation performance. By selecting points in the time-frequency domain, not only the signal-to-noise ratio (SNR) can be improved effectively, but the signal interference in different time-frequency domains can be isolated. The improvement of SNR makes TF-SSMUSIC have a more accurate DOA estimation than SSMUSIC in the case of low SNR. Especially, if source signals are separable in the time-frequency domain, TF-SSMUSIC can process them solely. In this way, the angle resolution and the number of predictable source signals can be improved greatly.

Optical-resolution photoacoustic microscopy (OR-PAM) has rapidly emerging as tool for label-free morphology and function imaging of the microvasculature in vivo with a high resolution. However, it is difficult to achieve real-time imaging due to the limitation of data acquisition time. Therefore, a sparse PAM (SPAM) has been proposed to obtain a high-resolution PAM image with relatively low sampling density. In order to successfully set up a SPAM system, the two key problems that we need to keep focus on are designation of the compressive sampling scheme and the corresponding image recovery algorithm. Typically, a random uniform sampling scheme is adopted. In this paper, a non-uniform sampling scheme based on low-rank matrix approximate is proposed to replace the conventional point-by-point scanning scheme to implement fast data acquisition. The effectiveness of the proposed non-uniform scanning scheme is validated using both numerical analysis and PAM experiments. As compliments for SPAM system, the total sampling points are dramatically decreased for a relatively high-resolution PAM vascular image and to implement accelerated data acquisition. Thus, OR-PAM is of great potential to find board biomedical applications in the pathophysiology studies of tumor and treatments for anti-angiogenesis.

In recent years, with the development of sensor technology and wireless communication technology, wireless sensor network (WSN) as the technology for information acquisition and processing is widely applied in many fields. It is important for nodes to know their localizations for further applications. In this article, a range-free localization algorithm in WSN that builds upon the Monte Carlo Localization (MCL) algorithm is proposed. It concentrates on improving the sampling efficiency by changing the weights of samples. More specifically, mobility is used to improve the sampling efficiency to make sure MCL can perform well even when the sample number is low.

In the UAV path planning study, due to the relative movement of multiple targets and the drone, long-term and large-scale UAV autonomous tracking has not been achieved. Therefore, aiming at this problem, this paper uses multiple moving target tracking algorithm to provide a real-time feedback on target position, estimates the later motion state of the target according to its position, and then performs the dynamic path planning by combining the feedback data and the state estimation result. Finally, The UAV path is optimized in real time. Experiments show that the proposed scheme can better plan the UAV path when multiple targets are in motion, thus improving the intelligence of the drone and the capability of long-time tracking.

In wireless sensor network localization, many anchor nodes and target node exchange information at specified time intervals to obtain the distance information between each anchor node and the target node. With this information, the coordinates of the target node can be achieved through the calculation of the positioning algorithm. However, as there are numerous negative factors like non-line-of-sight measurement, complex multipath fading, which leads to high-level localization error. To improve localization accuracy, an improved least square localization algorithm is proposed, which combines the least square localization method with the statistical filtering optimization strategy. The simulation results show that this algorithm can effectively reduce localization error and achieve more accurate localization.

Accurate and reliable distance information is essential to a variety of wireless applications, and ultra-wideband (UWB) signal can theoretically achieve centimeter-level ranging accuracy. In this paper, we design and implement an UWB-based anti-lose system, using commercial ScenSor DWM1000 module. The centimeter-level ranging accuracy of the system is shown under both line-of-sight (LoS) and non-line-of-sight (NLoS) experimental conditions.

Localization technology has been widely applied in various fields such as military investigation, natural disaster prevention, address search, and travel route planning. In order to guarantee the coverage range and localization performance of localization technology in both indoor and outdoor environments, research on indoor and outdoor seamless localization using global navigation satellite system (GNSS) and wireless local area network (WLAN) has attracted lots of attention. In this paper, a seamless localization method based on GNSS and WLAN is proposed. The method is able to switch smoothly from GNSS to WLAN localization in indoor and outdoor environments and outperforms either the GNSS localization or WLAN trilateration localization.

Land subsidence is a geological disaster caused by natural or human activities. The rate of change in early settlements is often extremely small and presents a challenge to monitoring. This experiment includes BeiDou positioning, multiple antenna, and high-precision baseline solution. It developed the BeiDou deformation monitoring system and used static relative positioning for high-precision land subsidence monitoring. We have adopted integrated hardware design, equipped with a variety of communication modules, satellite receivers, and embedded module in one. In addition, we have developed the corresponding communication protocol for data transmission. Finally, a corresponding monitoring interface software was designed on the client to intuitively reflect the settlement process in a graphical manner.

The mobility and feasibility of mobile ad hoc networks (MANETs) have to deeply rely on the accurate location information to support multiple applications. A wrong announced location of a node may cause some serious consequences. Thus, the localization and the location security should be considered as important parts of the whole design of the MANETs. In the traditional way, the location verification schemes need complex calculation and multi-step communication with base stations and other vehicles, which are strengthen the burden of the MANETs routings and increased the complexity of protocol. In this paper, an improved multi-layer location verification system (MLVS) based on the optimal common neighbor’s knowledge between claimer and verifier in MANETs is been proposed and discussed. In this system, each node in MANETs could have a trust value, and a mutually shared token scheme is provided to make the decision of the MLVS. Furthermore, the MLVS shows a reliable performance on the ability of attacker defense and accuracy in high node density networks.

With the progress of society and economic development and the expanding deployment scale of substations, the automatic monitoring of the foundation settlement in construction station becomes an important issue of the operation and maintenance in the power grid. BeiDou satellite navigation system is a global navigation and positioning system, which is independently developed to provide navigation, positioning, and timing services in China and its surrounding areas. This article describes the principles of BeiDou high-precision positioning. According to the land subsidence monitoring and warning requirements for substations, multi-antenna technology is introduced in order to solve the problem of the cost of large-scale land subsidence monitoring system. The monitoring interface software is designed to visually reflect the settlement situation. The monitoring system can effectively improve the monitoring efficiency and early warning capability of the ground subsidence in the power company’s substation. The system will also provide a basis for decision-making management and improve the reliability, safety, and stability of power grid operation based on high-precision field accuracy and trend.

With the development of artificial intelligence voice technology, array signal processing technology has been widely used in intelligent human–computer interaction. In this field, the sound source localization technology of the microphone array system is also one of the key technologies in the composition of intelligent systems, and it is also a research hotspot technology, which is of great significance for improving human–computer interaction ability. In this paper, the microphone array technology is used to demonstrate and improve the sound source localization of indoor speech signals. A spatial distributed microphone array localization algorithm is proposed to improve the accuracy of the indoor voice source signal localization in the presence of indoor noise and reverberation.

In the 5G communication network, D2D technology is considered to be one of the important components of the future [1]. D2D users bring multiplexing gain to the channel resources of multiplexed cellular users, improve system communication capacity, and reduce communication delay and terminal performance [2]. D2D also brings considerable interference to the reuse of cellular user channel resources [3]. The research on the resource allocation algorithm in D2D communication system mainly aims to reduce the interference between D2D users and cellular users. The improved bilateral rejection algorithm based on the best algorithm improves the stability and system performance of the whole communication system to a large extent [4].