With the development of computer science, more and more hardware implementations can be reproduced by software programming, bringing compact, cheap, and fast components to imaging instrumentation. In recent years, computational methods have been introduced into spectral detection, and computational spectrum acquisition implementations have emerged. This paper highlights the advantages of computational spectrum acquisition implementations by comparing them with traditional noncomputational methods. Then, focusing on the compact feature, we review the most representative implementations, and finally make discussion and offer an outlook.
In this review, we introduce some chip-based waveguide biosensing and imaging techniques, which significantly reduce the complexity of the entire system. These techniques use a well-confined evanescent field to interact with the surrounding materials and achieve high sensitivity sensing and high signal-to-noise ratio (SNR) super-resolution imaging. The fabrication process of these chips is simple and compatible with conventional semiconductor fabrication methods, allowing high-yield production. Combined with recently developed chip-based light sources, these techniques offer the possibility of biosensing and super-resolution imaging based on integrated circuits.
Crowd density estimation, in general, is a challenging task due to the large variation of head sizes in the crowds. Existing methods always use a multi-column convolutional neural network (MCNN) to adapt to this variation, which results in an average effect in areas with different densities and brings a lot of noise to the density map. To address this problem, we propose a new method called the segmentation-aware prior network (SAPNet), which generates a high-quality density map without noise based on a coarse head-segmentation map. SAPNet is composed of two networks, i.e., a foreground-segmentation convolutional neural network (FS-CNN) as the front end and a crowd-regression convolutional neural network (CR-CNN) as the back end. With only the single dot annotation, we generate the ground truth of segmentation masks in heads. Then, based on the ground truth, FS-CNN outputs a coarse head-segmentation map, which helps eliminate the noise in regions without people in the density map. By inputting the head-segmentation map generated by the front end, CR-CNN performs accurate crowd counting estimation and generates a high-quality density map. We demonstrate SAPNet on four datasets (i.e., ShanghaiTech, UCF-CC-50, WorldExpo’10, and UCSD), and show the state-of-the-art performances on ShanghaiTech part B and UCF-CC-50 datasets.
Chronic kidney disease (CKD) is a widespread renal disease throughout the world. Once it develops to the advanced stage, serious complications and high risk of death will follow. Hence, early screening is crucial for the treatment of CKD. Since ultrasonography has no side effects and enables radiologists to dynamically observe the morphology and pathological features of the kidney, it is commonly used for kidney examination. In this study, we propose a novel convolutional neural network (CNN) framework named the texture branch network to screen CKD based on ultrasound images. This introduces a texture branch into a typical CNN to extract and optimize texture features. The model can automatically generate texture features and deep features from input images, and use the fused information as the basis of classification. Furthermore, we train the base part of the network by means of transfer learning, and conduct experiments on a dataset with 226 ultrasound images. Experimental results demonstrate the effectiveness of the proposed approach, achieving an accuracy of 96.01% and a sensitivity of 99.44%.
The multi-objective optimization problem has been encountered in numerous fields such as high-speed train head shape design, overlapping community detection, power dispatch, and unmanned aerial vehicle formation. To address such issues, current approaches focus mainly on problems with regular Pareto front rather than solving the irregular Pareto front. Considering this situation, we propose a many-objective evolutionary algorithm based on decomposition with dynamic resource allocation (MaOEA/D-DRA) for irregular optimization. The proposed algorithm can dynamically allocate computing resources to different search areas according to different shapes of the problem’s Pareto front. An evolutionary population and an external archive are used in the search process, and information extracted from the external archive is used to guide the evolutionary population to different search regions. The evolutionary population evolves with the Tchebycheff approach to decompose a problem into several subproblems, and all the subproblems are optimized in a collaborative manner. The external archive is updated with the method of shift-based density estimation. The proposed algorithm is compared with five state-of-the-art many-objective evolutionary algorithms using a variety of test problems with irregular Pareto front. Experimental results show that the proposed algorithm outperforms these five algorithms with respect to convergence speed and diversity of population members. By comparison with the weighted-sum approach and penalty-based boundary intersection approach, there is an improvement in performance after integration of the Tchebycheff approach into the proposed algorithm.
Under the perspective projection assumption, non-Lambertian photometric stereo is a highly non-linear problem. In this study, we present an optimized framework for reconstructing the surface normal and depth with non-Lambertian reflection models under perspective projection. By decomposing the images into diffuse and specular components, we compute the surface normal and reflectance simultaneously. We also propose a variational formulation that is robust and useful for surface reconstruction. The experiments show that our method accurately reconstructs both the surface shape and reflectance of colorful objects with non-Lambertian surfaces.
The recommendation task with a textual corpus aims to model customer preferences from both user feedback and item textual descriptions. It is highly desirable to explore a very deep neural network to capture the complicated nonlinear preferences. However, training a deeper recommender is not as effortless as simply adding layers. A deeper recommender suffers from the gradient vanishing/exploding issue and cannot be easily trained by gradient-based methods. Moreover, textual descriptions probably contain noisy word sequences. Directly extracting feature vectors from them can harm the recommender’s performance. To overcome these difficulties, we propose a new recommendation method named the HighwAy recoMmender (HAM). HAM explores a highway mechanism to make gradient-based training methods stable. A multi-head attention mechanism is devised to automatically denoise textual information. Moreover, a block coordinate descent method is devised to train a deep neural recommender. Empirical studies show that the proposed method outperforms state-of-the-art methods significantly in terms of accuracy.
Traceability link recovery (TLR) is an important and costly software task that requires humans establish relationships between source and target artifact sets within the same project. Previous research has proposed to establish traceability links by machine learning approaches. However, current machine learning approaches cannot be well applied to projects without traceability information (links), because training an effective predictive model requires humans label too many traceability links. To save manpower, we propose a new TLR approach based on active learning (AL), which is called the AL-based approach. We evaluate the AL-based approach on seven commonly used traceability datasets and compare it with an information retrieval based approach and a state-ofthe-art machine learning approach. The results indicate that the AL-based approach outperforms the other two approaches in terms of F-score.
Smartphone, as a smart device with multiple built-in sensors, can be used for collecting information (e.g., vibration and location). In this paper, we propose an approach for using the smartphone as a sensing platform to obtain real-time data on vehicle acceleration, velocity, and location through the development of the corresponding application software and thereby achieve the green concept based monitoring of the track condition during subway rail transit. Field tests are conducted to verify the accuracy of smartphones in terms of the obtained data’s standard deviation (SD), Sperling index (SI), and International Organization for Standardization (ISO)-2631 weighted acceleration index (WAI). A vehicle-positioning method, together with the coordinate alignment algorithm for a Global Positioning System (GPS) free tunnel environment, is proposed. Using the time-domain integration method, the relationship between the longitudinal acceleration of a vehicle and the subway location is established, and the distance between adjacent stations of the subway is calculated and compared with the actual values. The effectiveness of the method is verified, and it is confirmed that this approach can be used in the GPS-free subway tunnel environment. It is also found that using the proposed vehicle-positioning method, the integral error of displacement of a single subway section can be controlled to within 5%. This study can make full use of smartphones and offer a smart and eco-friendly approach for human life in the field of intelligent transportation systems.
Permanent magnet synchronous motor (PMSM) has been widely used in position control applications. Its performance is not satisfactory due to internal uncertainties and external load disturbances. To enhance the control performance of PMSM systems, a new method that has fast response and good robustness is proposed in this study. First, a modified integral terminal sliding mode controller is developed, which has a fast-sliding surface and a continuous reaching law. Then, an extended state observer is applied to measure the internal and external disturbances. Therefore, the disturbances can be compensated for in a feedforward manner. Compared with other sliding mode methods, the proposed method has faster response and better robustness against system disturbances. In addition, the position tracking error can converge to zero in a finite time. Simulation and experimental results reveal that the proposed control method has fast response and good robustness, and enables high-precision control.
Long-term coherent integration can remarkably improve the ability of detection and motion parameter estimation of radar for maneuvering targets. However, the linear range migration, quadratic range migration (QRM), and Doppler frequency migration within the coherent processing interval seriously degrade the detection and estimation performance. Therefore, an efficient and noise-resistant coherent integration method based on location rotation transform (LRT) and non-uniform fast Fourier transform (NuFFT) is proposed. QRM is corrected by the second-order keystone transform. Using the relationship between the rotation angle and Doppler frequency, a novel phase compensation function is constructed. Motion parameters can be rapidly estimated by LRT and NuFFT. Compared with several representative algorithms, the proposed method achieves a nearly ideal detection performance with low computational cost. Finally, experiments based on measured radar data are conducted to verify the proposed algorithm.