Signal and Information Processing, Networking and Computers

Person re-identification remains a critical conundrum of intelligent video surveillance system, which intends to query pedestrians with the same identity. It is of great significance to extract features with sharp discrimination and robustness, especially for images captured in complex scenes, such as illumination variance, different pose and viewpoints. In this paper, we propose a normalized distance aggregation strategy, which integrates three discriminative pedestrian feature extraction models, i.e., low-level local maximal occurrence (LOMO), mid-level salient color names based color descriptor (SCNCD), and high-level feature fusion net (FFN). We utilize each feature descriptor to train two distance metric learning models, including large-scale similarity learning (LSSL) and cross-view quadratic discriminant analysis (XQDA), respectively. Accordingly, the ultimate distance is calculated as the sum of six optimized distance metrics by the min-max normalization with respective weights. In experiments, comprehensive evaluation results indicate that our framework gains superior performance with robustness to illumination, viewpoint as well as pose changes, and exceeds the existing methods on the public dataset VIPeR.

Recently, several attempts with Siamese Network in Similarity Learning has been made in visual tracking area, which gains a decent capability both in real-time and accuracy rate with simple networks. However, the potential of model pretraining in superior tracker performance and training efficiency has been constantly neglected. To address the issue in this regard, a novel loss function, namely imbalance-eliminate loss (Ie-loss) function, is proposed in this essay. Following and transforming the classic triplet thought in metric learning, our loss function can earn more training data and excavate the potential information among the triplets, which including positive and negative instances, as well as input exemplar. Further, we focus on eliminating the disequilibrium between positive and negative input by weighting elements in Imbalance-eliminate loss function. Additionally, we put forward a theoretical method about gradient to judge the validity of the triplet loss function. In experiments, our tracker with Imbalance-eliminate loss function achieves superior performance and attains state-of-art real-time standards in the multiple benchmark.

In the domain of human action classification, information obtained by optical equipment is used as the input source in most case. However, these optical devices are often affected by environmental factors, which bring some trouble to human motion recognition. The radar signal can avoid these situations, therefore the radar signal has been more and more significant part in this domain. In recent research, most researchers used the Micro-Doppler spectrogram of radar to classify the action. It only makes use of the time-frequency domain information, but ignores the spatial domain information. However, for the radar signal of the human motion, there should be some differences between the information in the time-frequency domain and the information in the other domain such as the space domain. In this case, we employed a CNN with two streams, which can use both Micro-Doppler spectrogram and the echo of the radar to recognize the action. In our architecture, there are two separate CNN processing streams for extracting the features of two stream of input, and the features are combined with a late network. The experiment proves that our method has better accuracy than the input of a single stream.

In massive multiple-input multiple-output (MIMO) systems, Pilot Contamination (PC) is considered as a key barrier in enhancing the Spectral Efficiency (SE) for fifth generation (5G) technology. To significantly suppress PC and to enhance the SE of the system a novel scheme “Efficient Sectorization-based Pilot Allocation (ES-PA)” is proposed in this paper. According to this ES-PA scheme the users (UEs) are divided on the basis of their signal-to-interference-plus-noise ratios into two zones i.e. cell center and edge zones. Furthermore, the edge zone is partitioned into N uniform sectors by using smart antenna techniques. The UEs in the cell center zone in the entire system are assigned with identical pilot sequences due to less Inter-Cell Interference (ICI). The edge zone UEs in each sector are assigned with mutually orthogonal pilot sequences due to the higher ICI and to mitigate the PC. The simulation outcomes reveal that the proposed scheme significantly reduces PC, attains lower mean squared error, higher data rate and better SE than conventional pilot allocation techniques.

License plate segmentation and recognition (LPSR) method is proposed for Chinese vehicles. The main process of LPSR is divided into three steps: vehicle license plate detection, character segmentation and character recognition. A simple algorithm, local binary pattern histogram (LBPH) is used for vehicle color license plate detection. For segmentation, connected components (CC) algorithm is used. The proposed algorithms have two advantages. First, if the Chinese character of a vehicle plate consists of only one connected stroke/part, the method segments the vehicle plate into required eight characters as Chinese standard license vehicle plate has eight characters (mostly one Chinese character, two/three English alphabets, three/four digits and a dot at third place). Secondly, if Chinese character has more than one stroke/part then this algorithm merges those strokes into one character. For recognition of license plate, support vector machine (SVM) is used. The results of proposed work demonstrate that license plate detection and segmentation methods perform better in terms of accuracy and performance and achieved accuracy rate of recognition is 96.7%.

The double threshold energy detector has attracted wide attention because of its low detection complexity. However, this comes at the expense of some sensing performance loss. Motivated by this, our paper proposes a novel double threshold detection method. Where maximum and minimum sensing threshold is redesigned and analyzed therein. Additionally, it is expanded to cooperative spectrum sensing for the issue of the hidden terminal. Simulation results verify the validity of the scheme.

In order to extract single-channel identical weak grating signal which is obliterated by noise during spacecraft ground test, a comprehensive extraction algorithm of the weak grating signal is proposed based on multiscale morphological filter (MMF), improved compete ensemble EMD with adaptive noise (improved CEEMDAN) and based on temporal predictability BSS (TPBSS). Firstly, the MMF is used to improve the signal-to-noise ratio, then the improved CEEMDAN is used to decompose the signals in multi-dimension. Meanwhile, the temporal predictability criterion is introduced to reduce the decomposition dimension and save the decomposition time. The virtual signal detection channel is constructed according to the results of decomposition. Finally, the TPBSS algorithm is used for signal-to-noise separation. The simulation results show that the algorithm can extract the weak grating signal well when it is obliterated by strong noise.

Effective prediction of crop pest species and quantities can provide a priori knowledge for pest control. However, there are many kinds of pests and the color and texture of different pests are similar, traditional image processing and recognition methods based on time or frequency domain features and classifiers are hard to meet the needs of high-precision recognition. In order to improve the recognition result of agricultural pests, based on the target detection algorithm Faster RCNN, we put forward a method for the classification and identification of pests on the basis of online hard sample mining and residual network. The main work is as follows: (1) According to the different size of pests, the detection ability of the network for small targets is enhanced by changing the size and number of anchors; (2) In the feature extraction stage, we use ResNet-50 with stronger feature extraction ability rather than the original VGG16 network; (3) In order to solve the imbalance of samples, online hard Sample Mining Strategy is introduced in the training stage. The experimental outcome demonstrate that our method is effective, the improved algorithm achieves about 5% improvement in accuracy, and the network has a stronger ability to detect small targets.

Feature extraction is an important part of machine learning models. However, traditional methods need professional knowledge about the specific field. We introduce an optimization method for feature extraction of radiation sources. The receivers get the sampling signals, and after applying Fourier transform, autoencoders train the data and output the encoded signals. Population based training (PBT) is used to optimize the hyperparameters. Compared to the traditional feature extraction methods, the proposed method achieve better result, the classification precision is above 95%.

Air pollution is the focus of attention. It is affected by time and space. At the same time, people’s attitudes toward air pollution are also worth noting. People with different pollutants may have different sensitivities. In order to compare this characteristic, this paper selects the historical data of air pollution in China’s Beijing-Tianjin-Hebei, Yangtze River Delta and Pearl River Delta cities for statistical and emotion analysis. Pollution in the Beijing-Tianjin-Hebei region is higher than other regions, but the improvement in the past few years is relatively high. The air quality in the Pearl River Delta is the best, but the improvement is low, and even there is a downward trend in some periods. The pollution in the spring and winter is generally serious. Then we use the ARIMA model to predict the future trend of various air pollutants, and find that although the overall air pollution shows a downward trend, especially PM2.5, PM10 is much reduced, but ozone has an upward trend and often becomes an important pollution in summer. In order to understand people’s attention to ozone pollution, we crawl the microblog data under the same pollution conditions of ozone and PM2.5 in Beijing for 6 days for emotional analysis, and find that people’s prevention and attention to ozone pollution is less than PM2.5.

In traditional supply chain systems, a centralized server is utilized to store information and a series of IoT devices are applied to manage and upload information. However, the server may be attacked easily and the IoT devices may forge the information, which result in a single point of failure, incorrect information problem and security challenges. To solve these problems, we design a distributed, secure and credible system by exploiting blockchain technology and game theory. The information in our system is consistent after consensus process, the stored information can’t be forged and the devices are motivated to behave honestly. The performance analysis demonstrates that our proposed system provides satisfied security and the numerical results illustrate that our award and punishment scheme can offer remarkable incentives for IoT devices to upload correct messages while keeping the system credible.

When a mixture of radiating sources and noises impinges on an antenna array, Direction-of-Arrival (DOA) estimation of source signals from the sensors array in presence of noise has attracted a lot of attention. Accurately estimating the DOAs of signals from the noisy observations is a challenge. With the development of sparse representation theory, atomic norm minimization (ANM) provides a super-resolution algorithm to denoise the observations of the sensors. Because of the convexity of the atomic norm, ANM problem can be solved by convex optimization, and DOAs of signals can be found from the dual polynomial of the prime problem. The problem is solved by Alternating Direction Method of Multipliers (ADMM) reasonably efficiently. Several simulations are conducted to verify the favorable performance of the proposed method by comparison with several existing methods such as MUSIC and Cadow’s methods, and the results demonstrate practical ability of the proposed method in both model order and DOA estimation in the presence of light noise.

Generative Adversarial Networks (GAN) is an unsupervised algorithm widely used in the field of computer vision. On the basis of traditional generative adversarial networks, a method combining deep Convolutional Neural Networks with Generative Adversarial Network for image processing is proposed. Kernel parameters of the initial convolution of the CNN network are initialized by using the Isotropic Sobel operator in the edge detection operator combined with the convolution kernels in multiple directions. We combined the CenterLoss and SoftmaxLoss as Loss function to improve the supervision. The algorithm can take the intra-class aggregation and class separation into account, making use of MINST and CIFAR-10 data sets to test. Results showed that the model we proposed had obvious advantages in rate of convergence and image recognition rate as compared with other traditional recognition algorithms.

Recently, the global air traffic has increased rapidly, most passengers choose to buy tickets in their own experiences. So accurately predicting flight ticket price is of great significance. Recently there exist some works on flight ticket price prediction. However, these works pay attention to solve a single flight price prediction. However, it’s unrealistic to build a lot of models for different flights. In this paper, we consider a lot of factors, including additional service data, flight data and time data. Firstly, we analyze data and extract flight information and time features. Furthermore, we propose a model: mask neural network, to solve flight ticket price prediction, which divides the features into two parts: individual information and general information. General information represents the portrait of a certain flight and individual information represents features of a certain flight. Since every flight has its own distribution, we cannot put individual information into the network directly. So we introduce mask concept and use general information as a mask to filter individual information. Experiments on a competition dataset with 3 months datasets introduce the effectiveness of our approaches, that can achieve the state-of-the-art in ticket price prediction.

Channel estimation is crucial in millimeter wave (mmWave) communication systems. The conflict between high angular resolution and heavy training overhead is a bottleneck to apply beam training-based channel estimation schemes such as hierarchical search and compressed sensing. In this paper, a novel two-stage adaptive high-resolution channel estimation (AHRCE) approach is proposed for mmWave system. In the first stage, hierarchical search based on normal-resolution codebooks is exploited to acquire the coarse estimated multipath components (MPCs). The number of measurements can be dynamically adjusted among training levels. In the second stage, a sparse signal recovery method to achieve high-resolution angles of departure (AoDs) and angles of arrival (AoAs) is introduced. The accurately estimated channel is estimated without quantization error by calculating a set of ratio metrics. Numerical results show that our proposed scheme achieves a more efficient tradeoff between the training overhead and mean square error (MSE) performance in broader SNR region than the other existing method.

Requirements engineering is a main question which spans across the whole software development lifecycle. Spacecraft’s software also needs to focus on software requirements, and manage the requirements across the entire software development lifecycle, including operation, maintenance, and the redevelopment. This paper summarizes six technologies and methods of requirement acquisition, including Interview, Questionnaires, Consult Strategic Objectives and Departmental Standards, Panel Discussions, Storyboard, Defect Tracking Management, analyzes different types of software requirement analysis combined with actual cases, Management software, Software for Scientific Research and production, Professional and Technical Software, these three types of software’s requirements engineering management. In the end, identifies some directions for future research. The aerospace company now more inclined to use all kinds of mature software platform based on a small amount of customization changes, succession advanced management concept in mature platform. It will improve the system stability, and at the same time, the flexibility of the excellent software platform can ensure the personal demands.

Research on helicon plasma propulsion mainly includes radio frequency (rf) power coupling, magnetic confinement, discharge mode. The great advantage of helicon plasma propulsion is its high plasma density. Radial velocity can be converted to axial velocity to enhance the thrust with a magnetic nozzle. The plasma density jumps during the plasma discharge is the vital judgment for helicon discharge mode. The article illustrates the apparatus of helicon plasma thruster and points out that there should be two plasma density jumps to verify the helicon discharging mode, and analyzes the plasma density collected by Langmuir probe. Meanwhile, the ion current of plasma flow is collected with retarded potential analyzer, and it clearly shows that there are two jumps in ion currents. The discharging images are taken for the capacitive discharging mode, inductive discharging mode and helicon discharging mode, respectively. Parameters for the discharging mode switch are acquired which will support the development of helicon plasma propulsion.

The antenna deployment process model which combines with the characteristics of the loop antenna structure setting the node equivalent modeling method, is established for a class of satellites with large loop antennas, is proposed to greatly reduce the dynamic dimension of the satellite attitude and orbit control of the loop antenna. The equivalent model satisfies the requirements.

The scheme of parallel propellant tank propulsion subsystem has been more applied on the large satellite gradually, and the adjustment of parallel tanks balanced consumption can increase the efficiency of propellant. In this paper, firstly, an analysis model about adjustment of parallel tanks balanced consumption is proposed. Secondly, some results are calculated under different parameters, and the influences of operating conditions such as the pressure, propellant mass and propellant volume rate on balanced consumption are obtained. Lastly, an experiment system about adjustment of parallel tanks balanced consumption is established, the balanced consumption is adjusted according to the proposed method. The effectiveness of the method is verified by the experiment. With the experimental results, the practical application suggestion about the proposed method is given. The research in this paper perfects the adjustment technology of parallel tanks balanced consumption, and improve the advantage of parallel tanks propulsion subsystem.

In this paper, according to the task requirement that the satellite control system is capable of accepting the replacement and expansion of the on-orbit module, the control system architecture of the data network based on the time trigger mode is studied. In this paper, a distributed architecture and a satellite control system with a component as an Orbital Replacement Unit (ORU) on-orbit are proposed, which makes the satellite control system more standardized and more reconfigurable and expandable while accepting on-orbit services. This paper analyses the key technologies of the satellite control system to achieve acceptable on-orbit services, and designs a demonstration and verification scheme to verify the timing scheduling strategy of the control system based on TTEthernet, the design of communication protocol, and the system reconstruction strategy after plug and play of intelligent nodes.

Shock response spectrum analysis is important for spacecraft instrument design, different Q values between given shock response spectrum and instruments may cause a considerable error. To get an appropriate shock response spectrum for a certain instrument, conversion is made for shock spectra with different Q values in this paper using hill-climbing optimization method. Consistency of the optimized shock spectrum and the specified shock spectrum is addressed in this paper, physical limitations of the shaker systems are not in consideration. Parameters are optimized according to the importance on the shock spectrum in sequence and the results shows that making optimization for amplitude can bring high precision. Runge-Kutta method with an appropriate time step is used to discrete approximation of single-degree-of-freedom (SDOF) system. Time history is synthesized using WAVSYN method and good consistency was obtained between the computed and required shock spectra. A given shock response spectrum with a Q value of 10 is converted to shock spectra with Q values of 1, 2.5, 5, 20 separately.

Helicon wave ionization source could provide abundant and excellent plasma for space electric propulsion. Now it is widely recommended and required to test and verify the technology. This paper designed and analyzed the structure of the helicon wave electric propulsion system and completed mechanics calculation with finite element method (FEM) software to validate the rationality and reliability. At last, it gave final conclusion and prospect.

Large scale strong magnetic field technology (namely magnetic sail) has important application value in space radiation protection and power propulsion. This paper mainly studies the key technology of large scale magnetic sail generation in low earth orbit (500 km, LEO). Based on the 3-D magnetohydrodynamics (MHD) numerical simulation model, the expansion phenomenon of satellite dipole field (Similar to magnetosphere, namely mini-magnetosphere) caused by plasma injection is studied. The numerical simulation results show that the dipole magnetic field carried by the satellite will continuously expand with the injection of plasma, and the attenuation of its magnetic field strength with distance will change from r−3 to r−2~r−1. In LEO magnetosphere region, its scale can expand more than 200 m. There are two main physical mechanisms for the expansion of the mini-magnetosphere: (1) the continuous enhancement of plasma hot pressing; and (2) the ring current structure formed by drift.

In the literature “An Autonomous Mission Planning Framework for the New Remote Sensing Satellites”, Mr. Xing separated satellites into 5 modes according to the satellite ability of planning tasks autonomously. NASA and ESA have launched satellites with the ability of mode 5 around about 2000. As for CASC, most of the remote sensing satellites in orbit do not have the ability of mode 4 up to now. In order to satisfy more and more complex user requirements, this article introduces an extensible autonomous task managing strategy, which is taking full consideration of the existing hardware foundation and suitable for multi-loading remote Sensing satellites. In fact, the strategy has been implemented with code, and has been demonstrated in a spacecraft still in development. Adopting this strategy, autonomous task managing ability of the in-orbit satellite can reach mode 4, that is, the satellite can arrange payloads working time by itself according to the tasks, without users considering collisions between each of them.

The remanence of the satellite interferes with the spaceborne magnetometers, which leads to the quality degradation of the measured space magnetic signals. In order to reduce the remanence interference in the signals, sources of the remanence are regarded as magnetic dipoles whose positions are fixed and magnetic moments change with time. An instantaneous linear mixing model was built to analyze the remanence interference in signals measured by the spaceborne magnetometers. Based on the independent component analysis algorithm, a method was designed to separate the independent components of the magnetic signals, and reconstruct the space magnetic signals excluding remanence interference. The performance of the method was verified by both the simulated magnetic signals and the in-orbit measured magnetic signals by the Tianzhou I cargo spacecraft. The result shows that the method can separate and eliminate the remanence interference in the signals measured by the spaceborne magnetometers, so that the quality of the signals can be improved.

Product quality has now become the key to the survival and development of manufacturing industry, and traceability of materials is particularly important. In order to meet the quality traceability requirement of UAV manufacturer in mass production mode, based on the analysis of the traceability problem in the actual production process of UAV, a method of real-time quality data acquisition and quality traceability of process based on bar code technology is proposed. On this basis, the design of UAV production line traceability system is carried out to trace materials. The traceability system has been successfully applied in an enterprise to realize real-time traceability and quality traceability of products. The research of this system is helpful to the digitalization and online of production process, and can improve the product quality according to the method of big data analysis.

In recent years, with the increase of single satellite performance, functional density and satellite life expectancy, micro satellite entered a rapid development stage. Satellite networking are urgently required for communication, remote sensing, navigation, ocean surveillance, scientific exploration and other fields. The construction of micro/nano satellite constellation group is rising day by day. Large-scale production of micro satellites has become an important research direction in the field of aerospace.Because of longer manufacturing cycle and higher cost, the traditional single star development and integrated manufacturing model is difficult to meet the mass production demand of the micro satellite constellation. In order to meet the mass production demand of micro/nano-satellite constellation, explore the batch development model of micro/nano-satellite constellation, this paper analyzed and proposed the overall framework and key elements of natch intelligent production line for micro/nano Satellite. Through the research on three-dimensional digital design, AIT process reengineering, production station planning, data management and control, intelligent equipment and other aspects, this article provides a construction idea for the subsequent batch and intelligent production line of micro/nano satellite.

In the background of 1.5 Gbps ultra-high code rate big data transmission, this paper proposes a parallel processing system for Multi-satellite test data processing of remote sensing load. From the two aspects of the hardware architecture of the system and the software logic of dynamic resource allocation, the specific design and implementation scheme of the Multi-satellite test data processing a parallel processing system for remote sensing load are expounded. That is to meet the needs of Multi-satellite normal testing, and save equipment development costs as much as possible, on-time monitoring of common equipment and resources to achieve a fast reconfigurable system architecture. The high-speed parallel processing solves the problem of on-time analysis of ultra-high code rate big data, and the single-channel parallel processing rate reaches 1.5 Gbps or more. The Multi-satellite testing is realized through the Multi-satellite sharing strategy, which fully satisfies the system-level verification test of researching and subsequent models.

As a new magnetic field measurement technology, magnetoresistive magnetometer has the absolute advantages in weight, power consumption and size, and plays an irreplaceable role in many fields. Because the data acquisition of the magnetoresistive magnetometer rely on ADC, So the paper presents the circuit design of a high-precision signal acquisition system based on sigma-delta ADC for magnetoresistive magnetometer. In this design, FPGA is adopted as the core controller of the system to control the read and write of six AD7710, it also completes receiving, caching, and sending ADC sampling data. The system self-test results show that the noise RMS value of the sampled data is 0.0019 mV when the sampling frequency is selected as 30 Hz, which is better than the typical output RMS noise value of the device.

With the development of electric propulsion technology, the measurement of plasma parameters has been widely used in the field of aerospace. In the process of electric propulsion performance examination and test, the plasma parameters are indispensable to the analysis of electric propulsion performance. Plasma parameters such as electron temperature, plasma density, plasma flux density play important roles in plasma technology, which are the important basis to determine plasma state of electric propulsion. There has a variety of diagnosis method for Plasma electric propulsion. But the radio frequency, electric and magnetic fields often existed in electric propulsion environment. Therefore, It is necessary to use the Retarding potential analyzer (RPA) and Langmuir probe (ALP) together to test the plasma parameters. Experiment data can be mutually verified to effectively improve the accuracy and reliability of test data. The test data of xenon and nitrogen plume obtained in this experiment is of great significance for the in-depth study of helical wave propulsion.

For the problem that 3D stereo target is costly and 2D planar target cannot cover entire field of view, it selects multi-sensor camera to capture image in the large field of view to improve defecting narrow data volume of single binocular stereo vision field. In order to improve accuracy of multi-sensor cameras, this paper selects marker feature points in the large field of view environment to draw virtual checkerboard target and multi-sensor camera captures image at a fixed position. Thus, virtual target under each camera orientation is large. A large target is formed in the calibration space of field of view and internal target of multi-sensor camera is calibrated by Zhang. Other virtual targets are separately calibrated to obtain the internal and external parameters of camera, and a certain camera coordinate system is set. For the absolute world coordinate system, the transformation matrix normalized to the absolute world coordinate system is calculated, and external parameters of camera are solved by Rodrigues parameter method. Through experimental verification, calibration accuracy error is around 0.2 pixel points.

Mars has become a hot spot in deep space exploration. New probe manner is required to investigate the planet more efficiently and flexibly. Mars aircraft is a new Mars exploration tool and the research in this field is of great significance. Quad-rotor aircraft is a key development direction of Mars aircraft research because of its simple structure, high reliability, the ability of hovering in the air and repeatable takeoff and landing. In this paper, the propulsion system, which is the most critical technology of the Mars quad-rotor UAV, is studied. When designing the propeller, the relationship between the blade inclination, rotation speed, radius and other factors on the blade lifting force is established, by using two-dimensional CFD simulation, and then the detailed structure of the propeller is designed. Based on design result, the carbon fiber propeller was manufactured and the lifting force measurement test and the load-bearing take-off test were carried out under the simulated Mars atmosphere environment. The results were consistent with the simulation analysis. This study can provide an important design basis for the development of Mars quad-rotor UAV.

During the process of GF-2 satellite assembling and testing, some equipment often needs to be disassembled and assembled repeatedly due to malfunction. After each re-installation, the measurement environment is not exactly the same, such as whether there is other equipment on the deck, which will affect the installation accuracy. Therefore, a precise measurement strategy based on reference cube network is designed during the GF-2 satellite assembly process. Each cube coordinate system of the equipment on the satellite is defined as a network node, then the reference cubes in the same measuring environment are connected into a network. Then a few hypothetical simplification conditions are introduced to link the network in different measurement environments to obtain the influence of measurement environments on the installation attitude of the equipment on the satellite, and to obtain a coordinated high-precision system for the installation attitude of the equipment, thus ensuring the on-orbit accuracy of the equipment.

The OneWeb constellation is taken as an example to study the feasibility of non-geostationary orbit (NGSO) broadband constellations used as relay stations for lower-orbit satellite’s communication with ground. Firstly, the technical route and frequency allocation of the OneWeb constellation are introduced. Then, benefits of LEO satellites communicating with broadband constellation as its client are given, together with analysis of differences between terminals of client-satellites and ground users. Finally, the link budget and dynamic characteristics of server-satellite to client-satellite are analyzed. The results show that with a link budget margin of 10.48 dB for forward link and 7.29 dB for reverse link, the need of reliable data transmission between two satellites can be satisfied. Dynamic characteristics analysis shows that the Doppler frequency change rate of both sides of the communication can reach up to 13.5 kHz/s, which puts higher requirements for tracking performance of client-satellite’s terminal than those currently in use. This paper also identifies some technical challenges to be overcome before the proposed scheme becoming operational.

This paper designs a vision navigation based steering wheel omnidirectional mobile platform in view of the large working space, low running efficiency and poor intelligence of traditional wheel type mobile platform. This platform is able to achieve the omnidirectional motion through the coordination of four steering wheel rotation and running motor by using the steering wheel of self steering ability as the drive unit, obtain location through visual acquisition of ground marking information and realize self navigation and positioning through, the displacement precision is less than ±5 mm and angle precision is less than ±0.8°, thus meeting the demand of automation, high precision and flexibility for spacecraft products and adapting to the development trend of automation and intelligence in the future.

SpaceWire fiber-optic Bus is a development of the SpaceWire ECSS-E-50-12C standard. Its transmission rate can reach 2 Gbps. However, it suffers from the problems of imprecise clock synchronization and link blocking in networks. Therefore, we analysis the problems and propose following methods. Firstly, we design clock synchronization algorithms to guarantee the precision of clock synchronization. Furthermore, we design static time distribution scheduling table which takes time-triggered traffic for real-time messages. Through simulation verification, the time precision can be within 58 ns while using clock synchronization algorithm. The use of scheduling table can effectively avoid network conflicts and reduce the delay of data transmission. The improved scheduling method can meet the real-time requirements of SpaceWire fiber-optic Bus systems, and will effectively promote the space application of fiber-optic technology.

When the noise covariance matrices in the filtering algorithm are not match with the practical system, the filtering performance may be degraded. To address this problem, a temporal difference learning Kalman filter (TDLKF) algorithm is presented for systems with noise covariance uncertainty. The key idea is to select appropriate noise parameters adaptively via trial-and-error interactions in an operational environment, such that the filtering performance is improved gradually. The adaptation scheme is implemented via temporal difference (TD) learning, which is an optimization-searching algorithm based on a value function updated iteratively as episode increases. The TDLKF is applied for an autonomous satellite constellation navigation system. The high performance of the algorithm is illustrated via simulation in comparison with the multiple-model adaptive estimation (MMAE) and the adaptive Kalman filter (AKF).

In order to improve the positioning accuracy of high resolution remote sensing satellite without ground control points, the accuracy of payload pointing determination must be improving. In this paper, the development of high accuracy payload pointing determination technology of satellite is introduced, and the main methods that achieve high accuracy are analyzed. There are two implementations, one is the highly stable structural design and the integrated configuration design which provides common installation reference for both attitude sensors and payloads, the other is measurement of the relative changes between the attitude sensors and the payloads in orbit and to achieve high precision payload pointing determination by compensation processing. Based on the investigation results, this paper proposes a pointing determination system design for multiple payloads, which realizes the high precision transmission from the measuring sensors to the pointing monitoring payloads from the system level, and can solve the pointing change caused by the internal changes of payloads such as the camera and LIDAR system. The design method can provide a reference for the future development of the satellite with high performance.

This paper puts forward a kind of self-excited RF power control system, which is more suitable for helicon plasma electric propulsion system than traditional RF generator based on a servo motor. Simple structure makes it smaller and lighter. It has no moving parts, and so that can be easily transformed into aerospace products. This system achieves the goal of impedance matching by using phase lock loop to trace load frequency. A verification circuit is designed, and the results prove that the design is feasible.

This paper mainly introduces the design scheme and the concrete realization method of small satellite launch site remote testing project. The paper analyzes the management process and technology realization of satellite remote testing from three aspects, namely the workflow, management mode and technology realization. The remote testing system solves the difficulties under the condition of long distance and narrow bandwidth, such as stability, security and flashover of remote control and multi-channel parallel transmission of audio and video data across the network segment. Through remote testing of launch site, it significantly saves manpower. It is conducive to providing technical support for the launch site with more back superior resources. With the overall popularization and application of launch site remote testing, the work interface is delivered to AIT, further promoting the separation of design and production as well as model development management mode transformation.

The accuracy and precision of data is the most important property of electrical measurement data in the assurance of aerospace parts. Based on the basic principle of measurement system analysis, we focus on the variance method to analyze the repeatability and the reproducibility of electrical measurement data. By studying the error of the measurement system to produce, we calculate the proportion of various error sources affect the measurement result, the interaction, synthetically determination of the measurement system status and measuring data quality, and adopt effective measures to reduce or eliminate the error. With the crystal oscillator measurement system using general instrument, we carried on the practice and its help us to realize the problem of data quality. By specifying the measurement interval clearly, the specific gravity calculation results of repeatability and reproducibility are reduced from more than 70% to less than 10% of the total variation.

In order to meet the requirements of the spacecraft mission, it is adopted usually that the spacecraft is made up of several cabins. In this paper, an application design of time-delay command which can make the spacecraft work steadily after separation is described. The centralized control before separation and self-control after separation are realized, which ensures that commands can be sent precisely in the process of separation. The feasibility method and working mode proposed in this paper have been applied in several spacecrafts, which improves the safety of spacecrafts, proves the stability of the system after the implementation of the delay instruction, and can be flexibly extended in subsequent missions.

Regarding an autonomous task planning problem of earth-observation-satellites, the author considers the combination of satellite task planning and action generation, and design eight kinds of actions including satellite attitude maneuver, load control, data memory, antenna pointing and data download. A lookahead algorithm based on expert knowledge is constructed and implemented. The algorithm calculates the task planning and action generation results that fulfil the requirements and restrictions through iteration within the specified lookahead step. The application in the project shows that, compared with the traditional task planning method based on using restrictions and on-orbit manipulation template, the method described in this paper can better fulfil the requirements and improve efficiency of satellite while reducing the amount of data injected by users.

This paper proposes a method using the deep neural network for satellite remote sensing image detection. This paper improves its network structure base on the YOLO-v3 network, which includes reshaping multi-scale feature maps, improving the network structure, improving the detection ability of weak targets in remote sensing images, reducing the number of convolution channels in CNN, and improves the detection speed. In this paper, the target database is established based on satellite remote sensing images such as GF-2 and ZY-3. The results of this paper show that the proposed algorithm has excellent performance in target detection accuracy and efficiency.

The measurement of atmospheric wind field is of great significance for global climate change, offshore storm detection, daily weather forecast accuracy and even national military security. Space-based coherent wind radar is widely concerned with its high temporal resolution, high spatial resolution, high sensitivity and small volume. The micro-vibration disturbance source exists in most spacecraft, which has a certain influence on the signal to noise ratio (SNR) of coherent wind lidar. Based on the analysis of the principle of coherent wind lidar system, the influence of satellite micro-vibration on SNR of coherent wind lidar is analyzed by taking a satellite micro-vibration data as sample. The experimental results show that the satellite micro-vibration has a certain influence on the SNR of the coherent wind lidar. While the mismatching angle is 3.77 μrad, the optimal aperture of the telescope decreases from 1.75 m in the ideal case to 0.45 m, The SNR is between −1.5 dB~1.36 dB, which is within the acceptable range. The research content of this paper can provide theoretical basis for the design of the satellite platform of coherent wind lidar.

Flexible and rapid computation of landing footprints for spacecraft of reentry phase is the crucial capability required for on-board options of possible landing sites. Mathematically, the problem is formulated as maximizing the crossrange for any downranges. Since solving the necessary optimality for the 3DOF model with Earth’s rotation proves to be a difficult task, some simplifications were made. Using energy-state approximations, the equations of motion are simplified. Based on the quasi-equilibrium glide condition (QEGC), the typical inequality constraints are enforced with velocity-dependent boundary constraints about the bank angle. Using Gauss pseudospectral method (GPM), the maximum crossrange optimization problem is converted to a nonlinear programming (NP) problem, which is solved by the sequential quadratic programming (SQP) algorithm. The approach is tested using the Apollo 11 model, and the simulation results prove the proposed algorithm could determinate landing footprints for the spacecraft of reentry phase reliably and simultaneously satisfy all the path constraints.

Thermal vacuum test is an important means of verifying the reliability of spaceborne units in space environments. Usually, the same type of spaceborn units are subjected to thermal vacuum test according to the specified test conditions before integrated into subsystem. This traditional thermal test methods cannot meet the requirements of Nano satellites in terms of timeliness and cost, which feature low cost and short manufacturing cycle. In order to reduce the cost under the premise of ensuring the quality, this paper proposes a overall thermal test method for GNC sub-system of Nano satellites to replace the original individual thermal tests. This paper first demonstrates the effectiveness of the thermal test, and then proposes the criteria for controlling the thermal test conditions of the sub-system based on the existing standard system. A satellite GNC sub-system is taken as an example to introduce its thermal vacuum test scheme. Test result shows that thermal mutual interference of DUTs can be eliminated by multi-zone temperature control and adopting different PID parameters for different kinds of units under test, even if their thermal response characteristics vary, which proves the validity of the test scheme designed.

In this paper, the protocol which is applied in the spacecraft distributed storage system is described. The spacecraft distributed storage system is made up with control modules and store modules which are connected through the SpaceWire network, and control modules can remotely configure the network and control store modules to finish file operations. All data transfer in the network should follow the protocol. Consider different application scenarios, the Serial Transfer Universal Protocol (STUP) and Remote Memory Access Protocol (RMAP) are used, and addressing mode takes the logical address and physical address. The system can provide services to users onboard through virtual addresses, and can be reconstruction without affecting the interface.

It is difficult to directly experiment validation verification of polar navigation elements in high latitude regions for commonage. The indirect grid navigation verification method based on simulated earth polar regions which sets an arbitrary point on earth equator as the fictitious pole is presented. Firstly, the transverse earth frame is defined, and then the transformation between the transverse earth frame, and the conventional earth frame is deduced, which lays the foundation for the polar region simulation. Secondly, transverse wander azimuth mechanical inertial navigation system under the transverse earth frame is established, which demonstrates characters of polar region in low latitudes, viz., the positioning and heading become indeterminate at the fictions pole. Thirdly, design transverse indirect grid navigation algorithm by referring polar indirect grid navigation algorithm thinking, till the virtual polar region environment and its corresponding polar navigation algorithm are possessed. Lastly, verify the reasonableness of the virtual polar region and its corresponding navigation algorithm through shipborne Inertial Measurement Unit crossed the equator.

The rapid variety disturb moment induced by large maneuvering payload will seriously decrease the precision and stability of satellite body. For the control requirement of precision pointing of two-axis rapid maneuvering load, the couple dynamic model of the satellite body and two-axis load was proposed, then the feedforward and feedback compound control scheme was designed to harmoniously control of satellite body and maneuvering load. Compared with the un-compound control, numerical simulation indicated that the compound control can effectively improve the precision and stability of the satellite body to 60% and better base will be provided for a rapid maneuvering load’s precision pointing.

Since spacecraft faces complicated mechanical environment which poses a great threat to the structure and sensitive equipment during flight missions, vibration measurement is always a focus for spacecraft design. In the recent remote sensing and space experimental missions, vibration environment during the launch period and micro-vibration environment while spacecraft operating in-orbit was measured and analyzed massively. A lot of flight data has been acquired.In this work, constructions, functions and technical index of monitoring and measurement systems installed on spacecraft are introduced. Applications of measured data during launch and in-orbit period are discussed respectively. Measured data during the launch period can be used to optimize mechanical testing conditions. Difference of the measured micro-acceleration between on ground and in-orbit is found. The measured data helps a lot at modification of mechanical testing conditions, adjustment of mechanical model and monitoring of micro-vibration environment. Future development of the measurement system is proposed in the end of this paper.

In order to meet the complex functional processing requirements of electronic equipment in the aviation and the aerospace fields, FPGA products are more and more widely used. By analyzing the technical characteristics of development for FPGA product in aviation and aerospace fields, comparing the effective methods of software engineering, this paper puts forward the method of implementing the engineering management of FPGA products by introducing FPGA verification procedure in the development for FPGA production. According to the different stages of the FPGA development process, the verification procedure can be divided into different technical processes, such as HDL Code Rule Checking, HDL Code Walkthrough, Simulation Verification, Timing Verification and so on. These verification procedures coverage from RTL code to gate-level net list. Which can effectively detect and repair the design defects of each stage in the FPGA product development. In one aviation project, many design defects are found and repaired by using this method, which effectively improves the quality and reliability of FPGA products.

In the uplink of deep space optical communications, a single-photon detector array is usually used on the flight optical terminal to receive data at the power level of several pW per square meter. The nested format of pulse position modulation and the inter-symbol guard times are usually used in this application. The timing information needed by the demodulator is also carried in this format. It is very crucial to obtain the timing information for demodulating and decoding the signal. An estimation scheme of timing offset for this modulation format is proposed in this article. This method is combined with the photon detecting technique of the up-down counter, providing synchronization as well as reducing the impact of the Earth and ambient background illumination. The method and algorithms for timing offset estimation are given and statistical models of signal processing algorithms are presented on the condition of the Earth background illumination.

To satisfy increasingly high precision pointing and stability requirements of future astronomical satellites and remote sensing satellites, and solve the contradiction between pointing control and vibration suppression, the paper studies a Stewart platform with abilities of both vibration suppression and pointing control. Firstly, the principle of integrated vibration isolation and pointing is explained, and the Theoretical transmissibility is analyzed. Then the dynamics model of the whole satellite with Stewart platform and flexible appendages is established. Based on this model, the simulation of transmissibility is done and the theoretical analysis results are verified. Finally, the vibration suppression and pointing control performance of the Stewart platform are analyzed by numerical simulation. The results show that the platform can meet the demand of vibration isolation and pointing adjustment at the same time.

On-orbit software injection function is a necessary means for on-orbit software maintenance of increasingly intelligent space-borne equipment. Based on the characteristics of the microprocessor BM3803, this paper presents a high reliability on-orbit software injection design method with strong error detection and correction, and self-locking prevention ability. Each bit of all of program data can be detected and corrected by bit-by-bit-two-out-of-three when it is wrong. The Bootstrap Programs achieve the purpose of loading conversion between memories with different bit widths, and EDAC. If the upgraded program cannot run, with the assistance of the FPGA the bootstrap programs can make the application return to the original program, so as to restore the bus communication link and basic functions. Experiments shows that this Design method can improve the reliability of system and make function expanding possible, also proves the feasibility of the scheme. The scheme has a good guiding significance for the on-orbit software maintenance design of space-borne equipment based on the microprocessor BM3803.

The structured, semi-structured and unstructured data of embedded aerospace software defects are distributed separately, and the big data mining, knowledge reasoning and aided decision making are impossible due to the lack of data association mapping and data clustering, making it difficult to use the software defect knowledge efficiently. In order to solve these problems, this paper proposes a method for constructing the knowledge graph on embedded aerospace software defects. This method obtains the software defect data by classifying the defect modes of embedded aerospace software and summarizing the fault analysis methods, so as to carry out researches on key technologies for knowledge graph construction, including knowledge modeling, knowledge extraction, knowledge fusion, knowledge storage and knowledge computing, and realize the applications of embedded aerospace software defect knowledge graph in aided code review, aided software defect prediction and localization, thus improving the efficiency of third-party software testing and evaluation as well as the quality and credibility of aerospace software products.

There is an urgent need to improve the stability of frequency sources in the field of communications. A quasi-linear analysis method for improving the frequency stability of quartz crystal oscillators is proposed. The groundbreaking factors of frequency stability are refined into linear factors, nonlinear factors and self factors, and derive an accurate calculation formula for each factor. Through the analysis and calculation of various types of crystal oscillator circuits, find out the factors of frequency stability, and give the optimization suggestions in the circuit design. In the frequency source circuit design, this method provides a fast and effective way to improve the performance of the frequency source.

Semiconductor components’ failure is one of the main causes for spacecraft malfunction. The quality control section has spent every effort to improve the quality of electronic parts and components by inline inspection and offline screening. However, there are products with unapparent failure mechanisms, which will slip through the quality control process and become a potential failure mode in application. The data collected in the semiconductor manufacturing process, including product design, material preparation, assembly, quality control, inception and screening, is highly nonlinear and multidimensional, making data mining an effective tool in processing industrial data, such as defects detection and fault diagnoses. Previous work has been done in improving the semiconductor quality based on data mining method, by specific functions like quality description, quality prediction and quality classification. In this work, the applications of data mining for semiconductor quality control were reviewed and the trends and challenges are analyzed.

With the development of communications technology, the requirement for the frequency source is lower phase noise, more highly pure spectrum and more highly frequency stabilization. In the microwave band, sampling phase-locked loop with high output frequency, low adding phase noise, small bulk and low power, and was widely utilized in some fields such as aviation military and satellite communications.The article through in-depth theoretical analysis to implement an X-band low phase noise sampling the phase-locked loop frequency synthesizer. The sampling phase-locked loop frequency synthesizer worked stably and had good capabilities. By using the ADS, we made the design simple. The design has a good reference value in the future.This paper takes 100 MHz crystal oscillator as the reference input, and finally outputs the carrier frequency of X-band by sampling phase-locked technology. The figure of phase noise is −112.71 dBc/Hz at 10 kHz offset carrier, clutter suppression reaches 75 dB and harmonic suppression reaches 40 dB, which achieves good performance.

In order to improve the signal transmission reliability of high-speed serial digital communication, this paper designs a digital signal transmission system based on the transceiver chip TLK2711. The scheme has the advantages of high transmission rate, simple interface and strong anti-interference ability. It can be widely used in various areas of digital communications. The TLK2711 transmits the 16-bit data and control code TKLSB and TKMSB through the 8B/10B encoder to generate 10-bit valid data to the 2:1 multiplexer, and then sends it to the serializer for completion. The data is converted from parallel to serial, correctly transmitted to the receiving end, and correctly parsed by the receiving device, and a reasonable transmission protocol needs to be defined. The article gives the detailed design scheme, usage and the working mode of each module of the system, examples and references are given for similar system design. In practical application, it can be improved according to the characteristics of the use plan and the actual needs of the design.

Space debris has a great threat to spacecraft safety. GEO fragments are difficult to measure by laser or radar due to the distance. Using two observation satellites to form a certain observation baseline can be used to determine the orbit of GEO space debris. The feasibility of observation is discussed, the positioning accuracy under different observational geometry conditions is estimated, and the search mode of the observation satellite and the working mode of the observation camera are proposed. For the orbit determination algorithm, the expressions of observations and related partial derivatives are given, and digital simulation of small inclination fragments and large inclination fragments are given. For the fragments with different inclination, the observation time of two satellites is different, the observation time of the small inclination fragments is longer, the geometric conditions are better, and the orbit determination accuracy is also slightly higher.

Optical cube is a common reference for the attitude determination of spacecraft equipment during assembly and integration. Three orthogonal normal directions to the reflective cube face can describe a reference coordinate system. Commonly, the cube attitude measurement often employs the use of theodolite autocollimation metrology. It requires that several theodolites be properly positioned in front of the cube to collimate the cube face and then to view each other’s collimated light. For some large and complex spacecraft, due to the layout of the instruments on the spacecraft, theodolites should be placed far away, sometimes the collimated light between theodolites may be blocked by other instruments or structure on spacecraft. We describe an alternative metrology for this demand. This metrology is based on gyro-theodolite which can provide a natural azimuth reference, so that theodolite bucking can be cancelled. This method was used to successfully measure the relationship information between in-cabin equipment and out-of-cabin equipment in the TianGong spaceship and has been used on other spacecraft projects. In this paper, the proposed metrology based on gyro-theodolites is compared with the conventional metrology based on theodolites in details, and the theory is discussed. Finally, a case application is given and the angular metrology results are analyzed.

The Chang’e-4 lander is the first probe which soft lands on the farside of the Moon, implements in-situ lunar surface exploration. As one of the key components, the telecommunication system is responsible for establishing the communication link between the lander and ground stations (or the relay satellite). Based on the characteristics of the Chang’e-4 lander, the paper analyzes the requirement of telecommunication system, proposes the optimal scheme, and presents the design method, implementation approach, ground verification and on-orbit results. The telecommunication system scheme takes the mission requirements into account, and inherits the design of the Chang’e-3 telecommunication system as far as possible, and carries out the system optimization design, which reduces the cost of the lander. Telecommunication system provides uninterrupted relay communication support for the lander, and realizes the reception of high sensitivity forward signal and high power dynamic lunar surface communication signal.

Three-axis magnetic sensor has been widely applied in magnetic measurements. However limited to the current technological level of processing and installing, the orthogonality of the three axes in magnetic sensors cannot be guaranteed, which will reduce the measuring accuracy. Therefore, calibration and compensation for magnetic sensors become necessary in engineering. Though some methods have come out to measure the errors of perpendicularity, some relatively rigorous device requirements or assumptions are required. Therefore, it is necessary to do some further researches on the calibrating and compensating method for the nonorthogonality of three-axis magnetic sensor according to the practical requirements. This paper proposes a new calibrating and compensating method for the nonorthogonality of three-axis magnetic sensor, which is free of the requirements on the magnetic field accuracy of the testing environment and the measuring accuracy of the sensor. Moreover, the real angles between two magnetic axes can be calculated by using accessible measurements, whose precision depends on the precision of the turntable.

The task planning plays an important role in mission management and satellite-ground cooperation. Considering the absence of overall design of task planning, we propose a novel framework of task planning based on precise simulation around earth imaging satellite. The system block diagram, data flow chart and main algorithms’ process diagram are presented here. The difference and conjunction between single and multiple satellite task planning are analyzed. The innovated results are able to be used to conduct the task planning software’s development and satellite development practice.

JF37 is a miniaturized high-reliability double-insurance separate connector. This paper focuses on the abnormal condition of mis-unlocking of separate connector under the mechanical vibration test on the ground, and proposes the improved method of the design including the design of cable combined part, locking spring, locking sleeve and etc. of the separate connector. The position of crimping, intension of string, stiffness of bending and the rock of the cable combined part under the vibration condition of the products are considered for the design of the cable combined part. The elasticity improvement of the spring is achieved by adjusting the length of sections and valid number of the circles of the spring. The strong pull-out force is decreased by converting the unified structure to distribution structure. The improved separate connector fulfills the requirement of mechanical and thermal test of the product. Meanwhile, it passes the test of mechanical test of deep space exploration on the ground and the test of separation simulation of exploration spacecraft. The proposed separate connector is with the characteristics of smaller size and lighter weight. Then, it saves the weight of MARS exploration spacecraft, and meets the requirement of MARS exploration spacecraft with smaller space and more difficult arrangement.

The intelligence of space control, especially the intelligence of docking tasks, is a trend. From the past performance of manual and automatic docking, the performance of automatic docking in terms of accuracy, smoothness and maneuverability is not as good as manual docking. The control strategy is extracted from the training data of the astronauts’ manual docking, and it is transformed into experience and knowledge by machine learning technology, integrated into the control law, improving accuracy and stability, improving the autonomy and precision of space rendezvous and docking, and achieving docking. Realize the controller to achieve the effect of people’s cognition, judgment and decision on the task. In addition, the complexity and diversity of spatial tasks put forward new requirements for docking technology. For new similar and different tasks, deep migration learning methods are used to realize online learning and final learning based on existing knowledge. The ability of the controller to be generalized to similar but different tasks is a direct manifestation of intelligence.

A satellite-borne equipment is capable of information exchange and telemetry acquisition and transmission. Due to the complex space environment in which the equipment is located, the prediction of the performance at the end of its 15-year life can provide good suggestions for the improvement of the product. Worst Case Analysis (WCA) is a reliability analysis method that takes into account of the device parameter drift in the circuit, and also can expose the most extreme state of circuit performance in the environment in advance. In this paper, a hybrid circuit simulation software Saber is used to analyze the telemetry acquisition circuit module of a satellite-borne equipment, and Monte Carlo analysis method is applied in the paper due to it is close to real result. We use Saber to find out the weak part of the circuit performance, and then give the improvement suggestions. At the end of this paper, we introduce some work that should be done in the future.

The high precision space time reference system is mainly used for space time foundation high precision modeling and calculation. Design in this paper, based on the space-time datum developed corresponding time benchmark and time system, coordinate system conversion and astrometry correction method, developed the basic algorithm library, include time system transformation, coordinate transformation and celestial bodies (stars and solar system objects) location calculation subroutine or function, and in a solar system objects, for example, in the imaginary observation position in the case of the moon, considering the gravity bending of the object, aberration and various disturbance factors, the calculation theory of the object position, and a simulation was carried out. This method can carry out high-precision modeling based on high-precision spatiotemporal datum, so as to improve the accuracy of geospatial information and provide theoretical basis for accurate geospatial information and other space applications such as space science and deep space exploration.

According to the characteristics of single small batch and the demand for the automatic installation of spacecraft fasteners, a set of automatic installation system for spacecraft fasteners is developed. This system can independently complete a series of actions, such as fastener grabbing, installation positioning, tightening, measuring force and so on. The screw separating mechanism realizes the single screw sorting and automatically conveys the screw to the installation position by compressed air blowing, the screw is tightened at the installation position by pneumatic sliding table, use electric torque gun to screw in and tighten with constant torque, through the electric torque gun controller to achieve the tightening torque data back collection. Automatic installation test was carried out and the actual tightening torque control precision met the requirements of the torque control precision of the spacecraft fastener installation.

In order to accurately measure the deviation between the actual size of aircraft components and the digital model, a new type of automatic measurement system is designed, which is used to measure the abnormal position that cannot be measured by traditional technical means. In view of the special features of aircraft component products, such as airfoil surface, three-dimensional space measurement and partial measurement path occlusion, the measurement system adopts visual system, trigger probe system, probe magnetic head replacement frame, and flexible design scheme of articulated mechanical arm. The measurement results can be directly output and the measurement report can be formed by comparing the test data with the real numerical model. This technology has the industrialization conditions and broad application prospects.

A method for autonomous imaging parameters’ adjustment based on solar altitude from remote sensing theory is proposed in this paper. The factors which influence imaging quality and the relationship between apparent radiance and solar altitude are firstly discussed; then integral times’ change under different roll angles in one orbital period and the internal links among solar altitude, roll angle, integral time and gain are analyzed; finally the best grading strategy are obtained and based on which the two-dimensional lookup table which can be used for the autonomous imaging parameters’ adjustment is built. Taking use of two-dimensional lookup table improves the image radiation quality and fairly reduces satellite’s control workload.

As the amount of data on satellite payloads soars, traditional data interfaces are no longer sufficient for engineering applications. The fiber optic interface has the advantages of high speed, small size, and long transmission, and will be used on satellites. Due to the extremely high speed of the fiber interface, a large amount of data is generated in a short time. If you use the first drop and then analyze, you need a large hard disk. This will result in a huge amount of test work and will not meet real-time testing requirements. The fiber detection system studied in this paper can perform real-time automatic detection on the onboard processor, improve the test coverage and test intensity of the onboard equipment, and ensure the reliability and stability of the onboard equipment to the utmost extent. At the same time, the system has the advantages of simple operation, easy expansion, low cost and convenient carrying. Through testing and verification, the system achieves the expected goal.

In order to evaluate and analyze the detection performance of lidar for optical properties in global ocean, the ability of the spaceborne lidar system to detect ocean water color profiles is evaluated by calculating the laser radar echo signal and the lidar detection signal to noise ratio. Monte Carlo method was used to calculate the signal-to-noise ratio, and the effect of layered water on the return signal was analyzed in combination with the Morley model. When other conditions were fixed, the greater the chlorophyll concentration of the water, the higher the intensity of the return signal. Analysis results show that the detection depth can reach 50 m, which can meet the measurement requirements of the coastal zone and the reef. By analyzing the influence of laser energy and receiving aperture on the depth of detection, the optimal design idea of the system is proposed, which can be used as a reference for system design.

Navigation receiver is an essential part of the satellite, with the gradual establishment of the Beidou Global Navigation System, navigation receivers add functionality of positioning and orbit determination based on BD3 navigation constellation instead of GPS and BD2 navigation constellation. In order to verify the performance of the novel navigation receiver in the comprehensive testing stage preferably, this paper designs a signal simulator which is suit for BD3 navigation constellation. This simulator not only realizes signal simulation of BD3 navigation signal under high dynamic circumstance by system design, but also updates modular by designing novel modular simulation models such as ionospheric delay module and Navigation message generation module. In addition, the simulator proposed by this paper has been already manufactured and the function of it has been verified through a joint testing with BD3 navigation receiver in this paper. It can be applied to the function and performance verification of BD3 navigation receiver.

Monopropellant propulsion systems with hydrazine are widely use for orbit and attitude control of remote-sensing satellite. Thermal design of the spacecraft propulsion system is very important to ensure its temperature in the range of 5–60 ℃. A thermal design method for remote-sensing satellite propulsion system is presented in the paper. Enhanced thermal insulation design is used to reduce the sensitivity of propulsion system to thermal environment. Heaters for the propulsion system are installed to prevent propellant from freezing. The heating power consumption of propellant lines and pipe valve is obtained through theoretical analysis. Taking remote sensing satellites in different descending node as examples, thermal design and on-orbit performance are discussed. The validity of thermal design for the propulsion system is proved by in-orbit data. The on-orbit thermal performance is well meet thermal requirements of propulsion systems. Impacts of propulsion system temperature are discussed.

PetroChina Southwest Pipeline Company (SPC) covers 70% of pipelines in mountainous areas. Owing to the complex topography and geomorphology, it is difficult to prevent and control the geological hazards. Traditional monitoring methods are difficult to achieve full coverage of pipelines. It is urgency to develop an economic and effective regional pipeline geological hazard monitoring system. In this paper, a new monitoring and early warning system for regional pipeline geological hazards is proposed. This system is described from the aspects of data fusion of GPS and InSAR, normalization of rainfall-deformation and combined warning model of heaven and Land. In order to verify the effectiveness and reliability of the system, the 50 km range of a pipeline is selected as the test area. The sentinel-1A, surface GPS monitoring station and rainfall quantitative monitoring station are used as data sources. In addition, the 82 potential risk areas found are monitored and warned for two years. The results verify the reliability of the system.

Peg-in-hole assembling structure is a commonly used mechanical structure for automatic assembly of robots in the aerospace field. With the structural constraints between pins and holes, the robot drives the installed equipment. With the force flexibility provided by the robot compliant control algorithm, the device will be installed in place along the guidance of the assembling pin. The traditional manual control method for pin-hole matching is more and more difficult to adapt to the installation of large-size equipment. To solve this problem, this paper proposes an improved force control method that takes into account the influence of contact force between the pegs and holes on impedance controller, and comprehensively considers the influence of elastic contact and friction between pegs and holes, so that it can adapt to the force compliance installation of large-size equipment, and the effectiveness of this method is proved by simulation experiment.

With the development of the space telescope in the earth and space observation, the aperture of the primary mirror has been increasing to 10 m order. The support of large aperture mirror affects directly the surface shape accuracy and stability of the mirror. The three main forms of mirror support, three-point support, hexapod support and whiffle-tree support are compared in terms of the number of support points and the applicability. The kinematics principle and the design considerations of the mirror mount structure design are discussed and the key techniques including optimization of the support points, structure athermal design and un-stressed assembly are scrutinized. This paper presents a systematical introduction of the support technique research process and is expected to provide design reference for related applications.

Machine learning has attracted tremendous attention in computer autonomous learning at present. Many intelligent applications can be realized by combining big data, machine learning and large computing power. In this paper, we present the basic concepts of machine learning, and explain the important characteristics and evaluation indexes. Then, introduce three typical algorithms of typical KNN, decision tree and random forest. Next, we show the basic principles of the three algorithms by describing the process. At last, we give a conclusion and show the future research of machine learning.

The power wireless private network is the basic information platform for grid informationization and intelligence. It fully penetrates and deepens the integration of the new generation power system. China’s 5G communication goes hand in hand in technical research, product development, standard promotion, etc., and receives strong support from the policy level. 5G changed the basic concept of 3G and 4G public communication networks before, and created a precedent for the coverage of public customers and industry customers. While expecting 5G, people should also remain calm, objective and rational, and can 5G meet the ubiquitous power Internet of Things demand in terms of technology maturity and construction progress? Does the 5G public network meet the key performance indicators such as security in the power core business such as bearer control? This article starts a basic exploration.

Recent years, with the rapidly development of aerospace remote sensing technology and commercial aerospace, the resolution of aerospace remote sensor has been improved continuously. Accordingly, it is possible to identify and track targets based on space remote sensing data. Furthermore, the corresponding application demands of various industries have also increased. In this paper, a new framework of target detection and automatic tracking technology based on remote sensing video data is put forward. Variation detection and the detection and tracking of moving targets are carried out with the data frame difference technology. In addition, the influences of channel detection on different frame tracking are analyzed in detail. Finally, the experiments of precise tracking for various remote sensing image data are also carried out. It is shown from the experiment that the proposed method can perform good detection effects. Moreover, as for the remote sensing video data, the problems of target tracking can also be efficiently solved, especially those for slow-moving targets and the target tracking with long frame intervals.

With the development of artificial intelligence, which has driven the revolution of the UAV (Unmanned Aerial Vehicles) industry. This paper introduces the development status of artificial intelligence technology in the field of UAV, and analyses the application of artificial intelligence technology in the field of UAV, Including the three aspects, intelligent perception, intelligent processing, intelligent decision-making, and proposed the development of intelligent UAS.

In view of the problem that the remote sensing image of UAV is greatly affected by the weather, this paper proposes a UAV remote sensing image enhancement technology based on improved GAN. Firstly, the noise feature is extracted by the noise feature extraction network based on convolutional neural network (CNN). Secondly, using the noise features extracted by CNNs, the generation of GAN for image enhancement is constructed. For the problem of large storage overhead, low computational efficiency and pixel block size limiting the size of the perceptual region, this paper introduces a full CNN model (FCN) as a generation network. Finally, a data set for training is constructed using image enhancement techniques based on damaged sample learning. The experimental results show that the enhanced algorithm in this paper shows excellent performance in remote sensing images of UAV.

Image quality improvement of remote satellite puts forward higher requirements for micro-vibration analysis and attenuation. From the engineering application viewpoint, the disturbance factors of data transmission antenna is identified, the analytical dynamic model of data transmission antenna is established, the theoretical identification method of model parameters is proposed combined with the experimental data of the equipment in this paper. Finally, the finite element - dynamic coupling model of antenna is established by hybrid modeling method, and the reasonability of theoretical model and finite element model is verified by comparing simulation results with experimental data. At the same time, the whole model provides effective input for subsequence satellite dynamic analysis.

The technological development and application of geosynchronous orbit (GSO) agile small satellite are researched. GSO small satellite has become a trend of future satellite development due to its application value. It has the agility of the research and development process, as well as the agility of control performance. In this paper, firstly the technology development status of GSO agile small satellites which have been launched into orbit are overviewed. Secondly, the application field and application value of GSO small agile satellites are analyzed. The differences and similarities among GSO agile small satellite, traditional large satellite and low earth orbit (LEO) small satellite are compared and analyzed. Then a propositional design scheme of a type of GSO agile small satellite for space surveillance and situational awareness is given and a new generation of electronic information architecture is selected for it. Finally, the Key technologies about control system of GSO agile small satellite are studied.

This paper establishes a custom index analysis system for big data of the mobile network and basic network information to solve the problem that the basic information cannot be quantified; an automatic inspection method and process for basic information based on MR grid and overlay business logic is established to solve the problem that large-scale network basic information cannot be quickly verified; also we establish an intelligent prediction for basic information and machine learning model of filling algorithm to solve the lacking and updating of basic information and provide a new way for basic information management of the service network.

In this paper, we present the first investigation about the similarity of APP usage by analyzing the data of hundreds of thousands of mobile users (MUs). Firstly, we define connection index (CI) and difference index (DI) to measure the degree of call relationship of two MUs and the degree of the difference of their APP usage, respectively. Secondly, we present the comparisons of the values of Pearson’s and Spearman’s correlation coefficients between MUs with call relationship (ReUs) and random selected MUs (RaUs), and explain the results of various APP types. Moreover, the changes of DI over CI for all kinds of APP are illustrated by observing a series of scatter diagrams. Finally, Z-test is adopted to confirm that the mean of DI for ReUs is lower than that for RaUs with a high probability. Through the above analysis, we could draw some conclusions that the call relationship has an impact on APP usage habit, and the APP usage patterns between ReUs have similarity.

As the rapidly development of 5G mobile communication network and smartphone market, almost every person takes a mobile phone during their daily time. To be sure, the expansion of smartphone makes people’s life more convenient. According to the latest research report, the current 40%–45% penetration of smartphone in the mobile phone market will reach to 65% by the year 2020. It is well known that the customers watch TV, chat with their friends and go shopping by their smartphones at any moment and anywhere. As a result, the little device carries abundant information of a person’s profile and behavior. In that situation, telecom operators could use data mining technology to develop personal daily routine information such as the busiest time in the day and which time to go to bed for rest. In that way, the telecom operators could take full advantage of this research achievement in accurate marketing. In this paper, a best-marketing time prediction algorithm (BMTA) is proposed to analyze the daily routine of customers and to predict best-marketing-time by using big data analytics. At last, the analysis results show that the algorithm proposed is high-efficiency.

With the rapid development of the Internet, the era of big data is coming. The weather forecast operation also is faced more challenges and opportunities. At present, the intelligent grid forecasting operation being carried out by the China Meteorological Administration will be based on meteorological big data, artificial intelligence methods and high-performance computing. This study uses the high-frequency grid multi-source observed fusion product generated by the HRCLDAS operational system of the National Meteorological Information Center as the grid observation data. The Sliding Two-Predictor regression correction method is used to test the full-grid point correction of the China’s regional ECMWF model direct output products of. And three different parallel computing schemes were compared with each other. The results show that the STPRC method can correct the ECMWF mode direct output product (DMO), and the grid MAE of 2-m temperature and 10-m wind speed can be greatly reduced. At the same time, the parallel computing scheme of GPU can greatly improve the efficiency of grid modeling. The combination of the two technologies finally produces high-accuracy, high-temporal and resolution grid forecasting products in a timely and efficient manner.

With the further development of the mobile network, telecom operators pay more and more attentions to end-to-end QoS (Quality of Service), like social application delay, in the focused scene. However, the conclusion based on the log tested at the scene or Key Performance Indicator (KPI) from the Operation and Maintenance Center (OMC) comes to inaccurate and inefficient. Our paper contributes a new method for analysis the end-to-end LTE network QoS in this area. Based on the operator’s big data, we firstly find out the “focused scene users” by using the identification algorithm, then we join the user side quality data in low latency, so we can analysis the end-to-end QoS of the scene we focused accurately and efficiently. Besides, it also powerfully supports optimization. The results also indicate that our new method is more efficient than the traditional ways.

Using neural networks (NNs) in post-processing ensemble prediction systems (EPS), it is important that reliable and accurate algorithm with a small computational expense should be provided in operational forecasting. Specific network architecture and choice of predictors as input nodes are designed to conduct approximation test for the calibration of forecast bias. To understand the strengths and weakness of different algorithm, six training algorithms have been compared with 7-15-1 network architecture. Results demonstrate that CGP had the best performance for a small computational expense in training and less forecast error. The same results can be drawn by statistical hypothesis tests. Furthermore, the elevation of station has significant impact on forecast bias. The skill score tends to be higher at the shortest lead time and is mostly ranging between 45% and 78% across all lead time for high-elevation stations. By comparison, the skill score of numerical optimization techniques is slightly higher, and has correspondence with the hypothesis testing results.

Video surveillance has been using in daily life widely and has significant impact in security field. Reading time information in video has becoming an essential part. Because of the complex background of timestamp and the high similarity between targets, the features extracted by traditional machine learning algorithm cannot meet the requirements and cannot achieve the required detection speed and accuracy. In this paper, we improved the algorithm with multi-scale features. Features extracted from convolution layers of different scales are used to predict, which could improve the recognition accuracy of small targets. In addition, because the size of the collected dataset is 2860 and the amount of data is not large enough, we use data augmentation and transfer learning to train. Pre-training on the public dataset SVHN firstly, and then trains on the special dataset. Experiments show that the algorithm can recognize timestamps accurately.

The development of big data technology provides more economical and effective methods for network optimization. Firstly, the platform logic architecture and the main functions of the network optimization platform based on rasterized big data are described. Then the key technical scheme of the platform is introduced. By rasterizing MR, the regional accuracy of the location problem is effectively improved, which is up to 50*50 m. By matching quality indicators and perception indicators to the grid presentation, we can build grid-level network, perception and value three-dimensional analysis capabilities, and more accurately combine perception and value dimensions to tract network optimization and planning. Finally, some application scenarios and cases are described to show the role of the platform in intelligent network optimization.

The analysis of “nodules” in lungs helps doctors to evaluate the situation and choose the appropriate treatments for the patients who are affected by lung cancer. In our study, we focus on the application of deep learning techniques for lung cancer diagnosis using CNN, DNN, and stacked autoencoders to enhance the accuracy and sensitivity of lung cancer diagnosis. In addition, we identified image preprocessing as well as deep learning preprocessing methods to extract the “nodules” found in lungs in order to classify cancerous lungs versus non-cancerous lungs.

The visibility predicted by the GRAPES-CUACE model and the objective forecast of visibility are corrected for this study. An improved correction method using the Machine learning for visibility predictions is established. Temporal and spatial variation characteristics of visibility before and after the correction are analyzed focusing on the applicability and improvement effect of the Machine learning. The error statistics and TS cores test shows that regardless of the CUACE model or the objective method, the visibility is generally lower than the observation, and the correlation coefficient between the CUACE model prediction and the observation is low. The visibility predictions effect of North China is better other regions, the corrected value is closer to the measured value, and the correlation coefficient is significantly improved. The correction method can improve the forecasting effect of visibility, mainly because the measured values and meteorological conditions of visibility are introduced, and the differences in meteorological conditions in different regions are considered, and the visibility of each model prediction is dynamically corrected in real time. However, there are still regional differences in the correction effects of different regions. In the future research, better design schemes will be adopted, and meteorological data assimilation methods will be used to reduce weather forecast errors to improve the correction effect.

NB network coverage assessment cannot be visualized through MR big data reported by user equipments, because Narrow-Band protocol does not support the Measurement Report (MR) function. NB network simulation based on LTE is used to visualize the whole network coverage rasterization assessment. This method could quickly locate the weak coverage areas in the early stage of business development, carry out targeted NB network pre-test work, reduce test manpower and time investment, also improve work efficiency of network optimization test. End-to-end optimization analysis get through the data interface of each layer from UE to service platform in IoT projects, which unify the data of each layer for analysis, and locate the problem accurately.

With the development of LTE networks, the scale of mobile communication networks has expanded rapidly, and equipment energy consumption has grown rapidly. This paper first introduces several existing LTE wireless power-saving technologies, and then uses various technologies to carry out single-station power-saving tests in the pilot area. By comparing the differences in base-station energy consumption and network performance indicators before and after the implementation of energy-saving measures, the impact of the technology on energy conservation and network performance was verified.

With the construction and development of information systems in various industries, data quality and interoperability have become more and more important. Data has been recognized as an important asset of an organization. In order to solve the “isolated information nodes” problem, data standards have been developed, among which information classification and coding are the most critical data standards. However, the data standards do not guarantee that information systems will be built in full accordance with the standards, thus affecting the interoperability between information systems. So the data conformance testing of information system has been paid more and more attention and applied in many industries. This paper analyses general theories and methodologies of conformance testing and standardization, summarize the existing problems for classification and coding information conformance testing. Then we put forward a new testing model for classification and coding information. Furthermore, we propose four measurements to evaluate the conformance results, namely, completeness, coverage, uniqueness and accuracy.

Traditional security service equipment is closely coupled with network topology, with fixed deployment location and complex management configuration. Security service is difficult to expand and upgrade, and cannot follow the changes of network application security. SDN-based security service chain (SSC) can realize security services on - demand orchestration and dynamic deployment. This article introduces the concept and classification of SSC, and expounds the architecture and creation method of SSC.

This study present a fast mode selection algorithm for a AVS2 encoder. At present, the traversal algorithm is adopted in the selection of prediction modes of the coding block in AVS2. Firstly the inter-prediction RDcost of the current block is calculated, and then the intra-prediction RDcost is calculated, the appropriate prediction mode is selected by comparing the RDcost. This scheme performs two prediction mode algorithms when encoding a block, it can select the appropriate prediction mode accurately, but it also generates a huge amount of computation. This proposal proposes a fast prediction mode selection algorithm based on spatio-temporal correlation of inter-prediction and intra-prediction blocks. According to the correlation of number and position of inter-prediction and intra-prediction blocks of adjacent frames or the same frame in the same video, we can quickly judge the prediction mode of the uncoded block, avoid inter-prediction or intra-prediction operations, and reduce the amount of computation and coding complexity. The experimental results show that this proposal can reduce the coding time by 2% for twelve universal test sequences, the total bit rate of the video is slightly increased, and the PSNR is slightly reduced.

With the Internet technology developing rapidly, network security has attracted more and more attention. Therefore, in order to protect private information against attack from malware, many people focus on the process of analyzing and recognizing raw traffic data to send an alarm to system in time and prevent damage. In this paper, we propose an improved method to recognize malicious network traffic data based on deep learning neural network, which applied convolutional neural network combined with Squeeze-and-Excitation Networks in order to learn spatial feature of network traffic data effectively and accurately.

With the popularity of mobile terminals and the increasing demand of video applications in mobile terminals, decoding speed of mobile terminal decoder becomes an important factor affecting video viewing perception. AVS is a series of standards formulated in accordance with international open rules to meet the needs of China’s audio and video industry. The software decoding speed of AVS2 (Second Generation of AVS) standard in mobile terminal is difficult to meet the requirements of real-time decoding of high-definition and ultra-high-definition video. In this paper, we study and implement a decoding optimization method using in Android – Selecting the sizes of $$ 16 \times 16 $$ and $$ 32 \times 32 $$ decoding units to reduce the computational complexity by intercepting certain frequency components during the procedure of IDCT (Inverse Discrete Cosine Transformation) at the decoder. With losing a small amount of image quality, this method can improve the decoding speed. Experiments show that this method can reduce the decoding time by 6%–9% in different resolution video.

With the increasing complexity of video coding and decoding technology, although the compression capacity of video has been greatly improved, there are still great challenges in meeting the real-time video coding and decoding requirements, especially 4K/8K real-time coding and decoding. In order to efficiently implement real-time coding and decoding, it is necessary to optimize the current coding and decoding technology. Most of the optimization schemes are based on the idea of stopping ahead of time and reducing the number of candidates. Some schemes optimize the codec according to the structure and principle of hardware. In this paper, various optimization methods of quadtree partition, intra and inter prediction and transformation, as well as some parallel pipeline optimization architectures for hardware and software implementation are reviewed.

With the development of mobile Internet, mobile operators are generally faced with the risk of network bottleneck and pipeline caused by the impact of data services. Among them, video service will become the most potential service in the future of wireless network, and bring unprecedented pressure and challenge to cellular network. For 2G/3G cellular networks that only support unicast, it is difficult for live video service based on wireless network to actually land due to bandwidth limitation. LTE eMBMS (enhanced Multimedia Broadcasting Multicast Services) solution has successfully realized the broadcasting capability in the whole network or multicast area. This article will introduce the system architecture, capacity and video services of eMBMS.

The data of original video is very large, so it needs to be compressed and coded to facilitate storage and transmission. With the continuous improvement of video resolution and diversification of content, video compression coding standards are facing new challenges. Although the current network bandwidth has been greatly improved, there are still great challenges for video transmission. As one of the core technologies of video coding, rate control plays the role of resource allocation and source output bit stream adjustment. It can optimize the allocation of the current remaining resources according to the channel state, and dynamically adjust the quantization parameters according to the feedback of the encoder, so as to provide the optimal video quality under the condition of limited bandwidth. In this paper, the current research on rate control is reviewed.

Rate control is an important part of video coding technology, which is used to control the encoder to output the code rate of a specific target and minimize the distortion of the coded video. Based on the rate control algorithm audio video coding standard 2.0 (AVS2), the fuzzy rate control has its limitations, and does not consider the correlation of video images. In this paper, an adjustment scheme of the quantization parameter (QP) is proposed. The QP of current encoding is adjusted according to the ratio of the inter-frame prediction block to the total coding block. First, when coding block division, to count the number and the pixels of the block using inter-frame prediction. Second, the final QP is adjusted according to the ratio of the statistical result to the total coding pixels. Finally, the newly obtained QP is substituted into the subsequent encoding process. The experimental results show that, under the common test condition, the new scheme can achieve significant Bjøntegaard delta bit rate (BD-Rate) gain and it is not difficult to achieve.

Person misalignment caused by detector errors or person pose variations is a critical problem in person re-identification researches. Without any alignment, background clutter and information ambiguity have a negative impact on feature matching across images. To address that problem, we propose an efficient deep neural network guided by human body region features for misaligned person re-identification. The proposed model we called person-region segmentation network (RSN) has a backbone for overall feature learning using deep convolutional neural network with global average pooling, and a branch for person-region feature learning employing human semantic parsing network. The effectiveness of the proposed RSN model can be verified via extensive evaluations on two public benchmarks, which contain Market-1501 dataset and DukeMTMC-reID dataset. Experiment results of RSN model reach Rank-1 of 90.67% and mAP of 76.59% on Market-1501, and reach 81.45% at Rank-1 and 68.34% at mAP on DukeMTMC-re-ID.

Face recognition has achieved extraordinary success recently due to the advancement of algorithms, technology, and hardware. Also, this is partly due to the richness of the face dataset. However, face recognition in complicated environment such as illumination, occlusion, pose is still a challenging task. Among them, illumination problems are still tough challenges in the field of face recognition. The datasets play an important role in deep learning, while there are few datasets specially for different illumination conditions and the currently existing face data sets have limitations. In this paper, we collect and set up a dataset called ILL-Dataset specifically for different illumination conditions. We also provide the detected dataset which is processed by using the face detection method. After it established, it can provide image data and promote the face recognition accuracy under various illumination conditions. The experiment results indicate the ILL-Dataset is a challenge dataset on illumination face recognition.