The Proceedings of the International Conference on Sensing and Imaging

Sensors and actuators are the basis of almost all electrical systems, especially if one views them as generalized inputs and outputs. But the approach to teaching them has been ad hoc, based almost entirely on an encyclopedic enumeration of their properties and applications. This approach is also evident in textbooks available. They are either specialized research monographs on a class of devices or are handbooks that describe many devices, sometimes exhaustively. A rational approach to teaching should be based on principles rather than on descriptions of functioning of the devices and should cover all or at least the main principles of sensing and actuation. But therein lies the problem. Sensors and actuators are based on numerous physical principles that cover all disciplines. To understand their operation and eventually to design sensors and sensor systems requires a broad knowledge in almost all areas of science and engineering. This requires a broadening of the base of instruction to include material science, physics, chemistry, and related subjects such as optics, nuclear engineering, fabrication, packaging, and others. Instead of specialization, the study of sensors and actuators favors diversification, and by so doing it stands in contrast with most other topics in the engineering curriculum.

Diffuse correlation spectroscopy and tomography (DCS and DCT) are emerging technologies used to noninvasive measurement or imaging of the blood flow index (BFI) at the microvasculature level. For both technologies, the reconstruction algorithm contributes greatly to the accuracy and robustness of BFI values as well as the detective sensitivity of diagnostic outcomes. In this article, we extensively reviewed the DCS/DCT reconstruction algorithms developed historically for extracting the BFI values and discussed their advantages/disadvantages. Furthermore, the DCS/DCT progress made in recent years to account for the tissue irregular geometry and heterogeneity is analyzed, and, through a computer simulation, we compared the performance of conventional and advanced algorithm in reconstruction of BFI values in human head. Finally, the future prospective of the DCS/DCT reconstruction algorithm and clinical implications are discussed.

The steady and transient flow simulations in an aorta model of normal subject were carried through computational fluid dynamic (CFD) technique. The steady- and transient-state computational fluid dynamic models of patient-specific aortic aneurysm were developed. The computed tomographic (CT) image data was used to generate the geometry of aortic models. The laminar flow was considered for simulating the flow of blood. The haemodynamic parameters like wall pressure, wall shear stress (WSS) and velocity distribution were estimated from the models. The obtained results depicted that the flow in the aorta model of normal subject was stable and aneurysmal aorta model became unstable. It is observed from the steady-state analysis that all the measured parameters from aneurysmal aorta model were higher than those obtained from the aorta model of normal subject. These measured parameters from this study could help the surgeons in assessing the severity of aortic aneurysms.

Blur is one of the most common distortions that affect image quality, and this work focuses on blur-specific no-reference image quality assessment (NR-IQA). Since various blur-specific NR-IQA methods have been proposed, we first give an overall classification of existing methods. Among all categories, we introduce 18 representative methods. Then, we conduct comparative experiments for the 13 representative methods with available codes on Gaussian blur images from TID2013 and realistic blur images from BID. Most existing methods have satisfactory performance on Gaussian blur images, but they fail to accurately estimate the image quality of realistic blur images. Therefore, it is needed to make further study in this field. At last, we provide discussions on realistic blur.

In this paper, we propose a novel quantity to measure the complexity of regions with inhomogeneous intensity in images. In order to describe real boundaries of objects, we further design an edge detector which is based on the similarity between object regions and those around them. Imbedding these two measurements of inhomogeneous regions into a level set framework, the proposed model is applied to segment HCC regions in CT images with promising results. Additionally, benefitting from the two measurements, segmentation is robust with respect to the initialization. Comparison results also confirm that the proposed method is more accurate than two well-known methods, the CV model and the BCS model, on segmenting objects with inhomogeneities.

Impurities (such as globular, blocky, or irregular solid) in pipeline easily lead to pipeline jamming, especially flowing fluid from main pipeline bifurcating into the thin pipeline. This situation may impact fluid speed and cause accidents. The novel computed tomography (CT) scanning mode is proposed to detect impurities in pipeline. Restricted by the scanning environment and impenetrability of the pipeline wall, this mode utilizes circular cone-beam scanning, and meanwhile it takes advantage of fluid flowing to obtain equivalent spiral cone-beam (ESCB) scanning mode, whose obtained projection data are affected by the pipeline wall. It treats the dynamic CT data as incomplete static CT data. Mechanical movement of the proposed scanning mode is simple and feasible, which can be easily adaptive to the live conditions of detected pipeline in service. According to the novel scanning mode, the corresponding approach which combines the general spiral cone-beam Feldkamp-Davis-Kress (FDK) image reconstruction algorithm and a local filter is presented to settle the truncated projection data. Numerical simulation tests show that the presented algorithm can facilitate relatively good local reconstruction images compared to general spiral cone-beam FDK.

This paper presents an improved linearity bootstrapped switch architecture for CMOS image sensor (CIS) application. Improper overcharging could lead the gate-source or gate-drain voltages of transistors to exceed the related supply voltage. As a result, transistors drive out their reliability limits. In this work, we report a dynamic bulk biasing circuit and a reliable clock doubling circuit for the purpose of improving the linearity performance of a bootstrapped switch in analog-to-digital converter (ADC). The bootstrapped switch is simulated in 65 nm triple well CMOS technology. The simulation results demonstrate that linearity with rail-to-rail swing can be achieved with an intended differential input signal. The signal-to-noise distortion ratio (SNDR) is 59.85 dB with input frequency of 21 MHz and sampling clock of 150 MHz, of which approaches to equivalent 9.6-bit linearity.

In this paper, we develop a novel algorithm to compute the union bound for maximum likelihood (ML) receivers for multi-input multi-output systems in real-time applications. The proposed strategy can effectively reduce the complexity of the minimum Euclidean distance search (MEDS) for the union bound by employing sphere decoding algorithms (SDAs) with modified Schnorr-Euchner (SE) enumeration. Furthermore, due to the importance of the initial radius (IR) for the complexity reduction of SDAs, the IR can be obtained according to channel statistics. Simulations illustrate that the proposed algorithm achieves considerable complexity reduction, while maintaining the accuracy of MEDS and obtaining an acceptable union bound. In other words, the proposed algorithm can be considered as a performance indicator, which can be exploited in many wireless communication algorithms, e.g., power allocation algorithms for cellular communications, precoding for multi-user MIMO systems, and dynamic pilot allocation for MIMO systems. With the aid of the performance indicator, the communication systems can be more flexible for different scenarios in the sense that the performance can be improved in terms of spectral efficiency, capacity, symbol error rate (SER), etc.

Recent experimental results have shown that the RF signal leakage is evitable due to the imperfect RF chain isolation. This paper focuses on the analysis of the impact of RF chain isolation on the self-interference cancelation (SIC) in co-time co-frequency full duplex (CCFD), considering multiple-antenna architecture. The idea of RF isolation impact analysis can also be applied to other fields, such as signal detection and signal processing. Using both simulation and experimental methods, the RF chain isolation impact on SIC in CCFD system is explicitly analyzed.

A new structure of a dual motor synchronous servo control system for precision tracking in the photoelectronic detection system is proposed. Compared with single motor direct control system, the antiresonance frequency of the dual motor control system is better than that of the single motor system. The current loop which can increase the low-frequency gain is used in the motor driver. The dual motor cross-coupling drive control with EtherCAT bus decreases the delays of the communication of the commands and the encoder feedback signals and increases the system bandwidth. The proposed servo control system is evaluated and compared experimentally with a traditional single motor servo control system on an industrial personal computer (IPC)-controlled dual-axis positioning system with two motor drivers. The experimental results show that the new structure of a dual motor synchronous servo control system remarkably reduces the tracking error. In addition, this new structure can be implemented easily on the real big telescope via the replacement of the motor drivers with higher power.

The time-of-flight (TOF) sensor can capture 3D surfaces very fast. However, it suffers from noises and low resolution, which limits its applications. To improve the resolution of TOF sensor, this paper proposes a novel method to fuse depth images and normal maps with complementary information. Different from traditional methods, which typically use photometric stereo (PS) to obtain normal maps, in our method, one extra light source is added to obtain normal maps via the Kinect sensor. Optimal depth images are then obtained by bilateral filter. Experimental results show that our method can produce more accurate depth images with smoother-looking object surfaces.

A flame temperature sensor based on using silicon nitride (SN) hot surface igniters (HSI) used as a dual-purpose sensor for both ignition and temperature measurement has been designed to measure the fuel-air equivalence ratio of premixed combustion systems. Equivalence ratio of premixed combustion systems provides a measure of efficiency of the system. Knowledge of the flame temperature along with the mass airflow rate of the air intake has been employed to calculate the equivalence ratio between 0.6 and 1, which is the range of equivalence ratios that most of the premixed systems run in practice. Different types of sensors from Kyocera and CoorsTek with different variations in the dimensions of the sensor element and the supporting ceramic element have been studied.

Eddy-current testing is an important nondestructive testing method in defect detection and evaluation of conductive materials. In testing of moving multilayer conductive plates, a rectangular exciting coil is placed perpendicular to the conductor, and cylindrical or rectangular coils are positioned below the exciting coil serving as pick-up coils. Accurate theoretical expressions of the induced voltage variations in the pick-up coils are derived, and the influences of the two pick-up coils’ exciting frequency and moving speed to the induced voltages are evaluated and compared. The rectangular and cylindrical pick-up coils are also used in testing for flaws in the conductors. The analytical calculation results are verified with experimental results.

Unscented Kalman filter (UKF) is a promising filtering method for nonlinear systems, but the statistical properties of the interference signals are not accurate, and the accumulated data error will cause error propagation. These factors will lead to the results of UKF inaccuracy, even divergence. This paper presents an improved UKF method based on the covariance intersection algorithm (CIA). According to the real value and the filter value, the improved value is given by this method. Meanwhile, the computation about the correlation information between the real value and the filter value is avoided by the CIA. Compared to the conventional UKF, the presented method can significantly outperform the conventional one and easily improve the system accuracy. In addition, Kalman filter is applied to linear systems, and the accuracy of Kalman filter is also further improved by this method. The simulation shows the effectiveness and superiority of this method.

Aiming at the existing problems that the image denoising algorithm based on integer-order partial differential equation could lost part of edge and texture information. This image denoising algorithm based on fractional variational theory was proposed by the theory of fractional calculus and partial differential equation. The denoising model proposed in this paper introduces and implements the numeric computation of the fractional variation partial differential equations by constructing the fractional differential mask operators along eight directions of image. The simulation data prove that the image denoising algorithm based on fractional variation theory compared with the traditional image denoising algorithm could better retain the edge and texture detail information, obtain visual effect, and properly improve the signal-to-noise ratio.

The total variation (TV) model has been used for removing the mixed additive and multiplicative noise. However, the restored images inevitably suffer from the staircase artifacts. In order to overcome this disadvantage, we propose two new variational models by combining the TV with overlapping group sparsity. Then the alternating direction method of multiplier (ADMM) is applied to solve the proposed models. Numerical experiments demonstrate that our methods are competitive with the state-of-the-art methods in visual and quantitative measures.

Target images of range-gated imaging in adverse turbid environments still suffer from the degradation of inhomogeneous turbid medium over the laser transmission path. Based on longitudinal laser tomography, a novel image restoration method is proposed to remove the interferences of the inhomogeneous turbid medium from the degraded target images. The degradation caused by the turbid medium is approximately estimated, assisted by some prior system parameters and the backscattering images of the turbid medium, which can further be used for removal of the interferences of the turbid medium through a proper total variation model. Experimental results demonstrate that the proposed image restoration method can effectively eliminate the interferences of inhomogeneous turbid medium and achieve exactly the reflectivity distribution for the target behind the turbid medium layer.

Object detection in natural images is evolving, with enormous commercial achievements, becoming relatively common in every industry. Modern research in this area is progressing in many directions, with numerous different techniques being proposed to achieve state-of-the-art detection performance. Recent object detection methods use two steps to detect high-quality objects: first, it generates a set of object proposals as accurate as possible, and then these proposals are passed to object classifier for post-classification. This paper presents an efficient new hybrid object proposal method, which gets the initial proposal by computing multiple hierarchical segmentations using super pixels and then ranks the proposal according to region score – which is defined as number of contours wholly enclosed in the proposed region, passing only the top object proposal for the post-classification. Passing few object proposals in the object detection pipeline for post-classification speeds up the object detection process. This paper demonstrates that our method results in high-quality class-independent object locations, with mean average best overlap of 0.833 at 1500 locations, resulting in a superior detection rate in object detection tasks at relatively fast speeds – as compared to object detection methods using selective search – and greatly reduces the false-positive rate.

To overcome the resolution limitation of magnetic resonance imaging (MRI), performing super-resolution (SR) on these clinical images is needed. Recently, a high-order regression-based SR framework demonstrates its advantage in producing fine details for MR image. However, the high time cost limits its application. To reduce time complexity, an adaptive-order strategy is proposed in this paper. Image structure tensor is used to classify the whole voxels into different groups. Afterward, the regression order is adaptively selected according to the classification result. Numerical experiments demonstrated that the proposed SR method can get a good balance between reconstruction quality and computation efficiency.

Sparse sampling can reduce the total radiation dose received by patients in the process of CT imaging. But in this situation, the reconstruction images can be severely degraded by strip artifacts. MAP algorithm with non-local prior has been applied to two-dimensional CT reconstruction from sparse-view data and generates high-quality image. However, the applications of non-local method in three-dimensional cone-beam CT are limited by its massive calculation. In order to remove streak artifacts and preserve detail information better, with the help of CUDA, we introduce non-local model into the sparse scan imaging of CBCT. The experimental results show that better reconstruction images can be obtained by the non-local prior in terms of the subjective visual effect and the objective evaluation indices such as the peak signal-to-noise ratio and the structural similarity index.

In recent years, binary keypoint descriptors have become ubiquitous in CV applications for their efficiency in computation. However, existing binary keypoint descriptors still face the problem of being less robust and discriminative. To tackle this problem, this paper presents a binary keypoint descriptor based on a newly proposed local pattern named local hierarchical octagon pattern (LHOP). The LHOP descriptor is much faster than SURF and ORB descriptors by creatively combing a newly designed orientation estimation method and the slanted integral image. Compared to the state-of-the-art keypoint descriptors, the main features of LHOP descriptor can be highlighted as follows: (1) It is robust. (2) It is efficient to compute. (3) It is highly discriminative. (4) It is memory saving and compact. Experimental results demonstrate that LHOP descriptor is at least 102 times faster than SIFT descriptor under almost the same matching performance. Moreover, the LHOP descriptor offers the better matching performance compared to other binary descriptors.

The needs of high-precision earth observation have led to the development of high-resolution and high-dimensionality RS data and greatly promoted the standard for processing and application of airborne hyperspectral images. The varying brightness gradients of the airborne images cause problems in generating “seamless” mosaic for hyperspectral surveys, which severely affect the radiometric consistencies for subsequent analyses. We present a semiempirical method to generate seamless mosaicking of multi-strip airborne hyperspectral images and introduce the model principle as well as the calculation process in detail. The experimental results based on HyMap images in Lop Nor area show that this method can efficiently remove the illumination gradient in both single image and between multi-scene images. Moreover, the MNF-transformed images and spectrum from overlap were chosen to assess the model; the results show that the Hapke-based model can be used to improve the airborne hyperspectral mosaicking effect and have great potential to subsequent quantitative applications.

Large field of view (FOV) imaging with high spatial resolution has been increasingly required for numerous applications in recent years. Obviously, conventional photosensitive detector with tens of megapixels cannot satisfy the requirement. As a result, gigapixel cameras based on the multi-aperture imaging have become a possible solution to overcome the above limitation. In this paper, we developed an alternative gigapixel imaging system which implements the multiple CMOS chips mosaic in the external optical path and presented the computation methods for calibrating the vignetting distributions and other geometric parameters in the system. Consequently, our gigapixel imaging system has achieved the performance of 24 Hz, 0.2Giga, single-pixel resolution.

Airplane engine is one of the most important parts in the plane, and the rotating parts are the key parts of the engine. In order to monitor the real-time statue of the rotating parts of the engine, a novel in situ airplane engine computed tomography system which takes advantage of the objects’ self-rotation has been proposed. Some featured artifacts exist due to the existence of static parts of the engine, causing difficulties in reconstructing the object. In this paper, these artifacts are introduced and discussed, a novel method based on expected value correction is put forward to eliminate them, and numerical experiments are taken to verify the effects of the method.

Low-dose CT is an effective solution to alleviate radiation risk to patients; it also introduces additional noise and streak artifacts. In order to maintain a high image quality for low-dose scanned CT data, we propose a post-processing method based on deep learning and using 2D residual convolutional networks. Experimental results and comparisons with other competing methods show that the proposed approach can effectively reduce the low-dose noise and artifacts while preserving tissue details. Factors that may influence the model performance, such as model width, depth, and dropout, are also examined.

With the development of computerized tomography, nondestructive testing technology has been widely utilized in the industry. In this paper, we try to extract the cracks contained in certain kind of workpieces from the industry, so as to predict their usability and left lifetime. The cracks are in tubular form in two-dimensional images, and many traditional tubular structure extraction methods show various difficulties in real applications. The main contribution of this paper is that an approach combining the phase congruency (PC) and phase symmetry is proposed to extract the tubular structures. Experiments on a kind of 3D workpieces from the industry against other popular methods are performed to verify the effectiveness of the proposed method.

Non-rigid multi-modal image registration plays an important role in many medical applications. Many advanced methods have been developed in deformable registration. However, conventional non-rigid registration assumes a global smooth deformation field throughout the image, so difficulties arise with handling the discontinuous displacement field and lead to poor correspondences at these sliding boundaries. These discontinuities exist for organs such as liver or lungs, which have a sliding motion during respiration. In this chapter we use total variation (TV) as a term to preserve the discontinuous boundaries for liver registration on computed tomography (CT)/magnetic resonance imaging. With benefits from the parametric transformation model known as free-form deformation, we analytically acquire an explicit optimization scheme for our method and compare our method with L2 regularization on both public and clinical data sets. The proposed method has been demonstrated to have a more credible displacement field near the discontinuous interface in both the public 4-dimensional CT data set and the clinical CT/magnetic resonance imaging data set.

牋Fingerprint enhancement and orientation field estimation are two of the core algorithms of fingerprint preprocessing. Performance of existing methods is usually poor for low-quality fingerprint images. This paper proposes to use a directional diffusion filter bank so that any local region of a fingerprint image can be always effectively enhanced by a filter of an optimal direction. A final enhanced image is obtained by selecting optimally enhanced pixels from the filter bank according to a local quality measurement based on a spectrum analysis, and at the same time, an orientation field is given by the selected filter for each pixel. Experiments show that the algorithm is superior to the existing methods and robust for images of poor quality.

RFID has been widely used in monitoring applications. In healthcare applications, events have to be detected in real time to make decisions. In this paper, we study optimization algorithms of event processing in RFID-enabled healthcare monitoring applications. We utilize non-deterministic automata (NFA) to model event processing. To reduce partial matches in event processing, we take advantage of a special data structure to maintain the events in memory. Event detection is accelerated by introducing context information as context can be used to delete many running instances which would not generate output complex event. Experiment results show that our methods are efficient and sound.

Wurster tube plays an important role in the circulation of solid particles in a fluidized bed coating process. To investigate the flow hydrodynamic behaviors inside a Wurster-type fluidized bed, a unique 3D electrical capacitance tomography (ECT) sensor is designed and constructed to visualize the gas-solid flow structure in a Wurster fluidized bed. The ECT sensor consists of 16 electrodes, and the electrodes are attached on the inner wall of the tube. Several “cold” tests are carried out by changing the fluidization air rate and the gap height between the Wurster tube and air distributor. The solid distributions in three dimensions are reconstructed based on the sensitivity maps and measured capacitance vector. Key parameters including the measured capacitances and averaged solid concentration with different operating conditions are given and discussed. The experimental result gives valuable information for the optimization of the fluidized bed coating process.

This paper presents preliminary results of imaging the flow on a thick pipe with electrical capacitance tomography (ECT). As a contrast, the images on thin pipe are also given. The study is based on numerical simulation with COMSOL and MATLAB. The pipe thickness of the ECT sensor is 2 and 12 mm, respectively. The capacitances and sensitivity maps are analyzed before image reconstruction. Flow patterns used in the study include annular flow, bubbly flow, core flow, and stratified flow. To improve the quality of image reconstructed in the thick pipe, the capacitances of adjacent electrodes are eliminated. At the end of the paper, analysis is conducted in terms of relative image error for different flow patterns.

X-ray phase-contrast imaging has attracted much more attention in researches, due to the better imaging contrast than absorption when the sample is made of light elements. However, X-ray phase retrieval procedure based on differential phase-contrast imaging (DPCI) is time-consuming because of the phase stepping technique, which is a necessary step of accurate retrieval of phase information. In this work, we proposed a single-shot method based on DPCI without phase stepping by designing a new absorption grating. It can successfully distinguish the characteristic of the sample with greatly reduced imaging time and radiation dose, which makes it more practical in medical and industrial applications. An experimental data is implemented to support our idea.

This paper presents an automatic method for liver tumor segmentation in CT scans. The proposed segmentation algorithm is a two-stage process. In the first step, the curvature filter is employed for removing the noise in CT images, and a trained mask is used to be a spatial regularization to constrain our segmentation in a specific region. In the second step, basing on the preprocessed results, we combine random forest with fuzzy clustering to segment liver tumor. In the experiments, the proposed method obtains promising results on the liver tumor segmentation challenge testing dataset. The calculated mean scores of Dice, volume of overlap error (VOE), relative volume difference (RVD), average symmetric surface distance (ASD), and maximum symmetric surface distance (MSD) are 0.47, 0.65, −0.35, 11.49, and 64.31, respectively.

Light sheet fluorescence microscope (LSFM) has become one of the most promising 3D microscopy techniques due to its prominent lateral section illumination. In addition, comparing with the confocal microscope, it has low phototoxicity and high temporal resolution. In this paper, we proposed an easy-to-operate tiling light sheet fluorescence microscope, which integrates a cylindrical lens, a galvanometric mirror, and an excitation objective to produce a light sheet with desired numerical aperture and axial resolution and utilizes an electrically tunable lens (ETL) to enlarge the limited field of view (FOV) of single excitation section following the temporal multiplexing. Imaging performances have been characterized from the experiment by imaging the 20 nm fluorescent beads with water immersion, and the benefits of our proposed technique have been shown in volumetric resolution, FOV, and speed as well as contrast.