Search

Synthetic aperture radar automatic target recognition (SAR ATR) has ushered in a new era dominated by deep-learning (DL) techniques. However, the DL-based recognition systems inevitably confront catastrophic forgetting for learned knowledge and overfitting for the new, once deployed in openly dynamic scenarios where targets of new classes continually appear with few-shot instances. For practical applications, a decoupled self-supervised subspace classifier with few-shot class-incremental learning (FSCIL) ability is proposed for prompt knowledge transferring and stable discrimination, w.r.t., intrinsic and domain-specific challenges of the FSCIL of SAR ATR. Specifically, observing the significant componentity and azimuth sensitivity of targets in SAR imagery, two self-supervised tasks powered by a scattering mixup module and a rotation-aware transformation module are designed to synthesize virtual samples and related labels to unleash the classifier's transferability to future categories while enhancing its discriminability to fine-grained scattering patterns. Once deployed, the model's parameters are frozen to decoupled with dynamic worlds for general knowledge extraction. At inference, a subspace classifier with class-aware target priors proposed by a max-coverage feature selection mechanism is formed for stable point-to-space discrimination. Extensive experiments on three FSCIL datasets built from SAR-AIRcraft-1.0, Self-owned, and MSTAR datasets, which cover various categories captured by airborne and spaceborne SAR payloads, show the state-of-the-art performance achieved by our method compared to numerous latest benchmarks.

Near-surface air temperature is a crucial weather parameter that significantly impacts human health and is widely utilized in numerical weather forecasting and climate prediction studies. However, the most common ground-based meteorological station observation and radar observation are often limited by geographic and natural constraints. With the advantages of global coverage and high spatiotemporal resolution, satellite remote sensing has become a valuable support in overcoming data scarcity issues related to ground-based station and radar observations in complex geographic and natural conditions. Although remote sensing indirectly reflects atmosphere variables (e.g., near-surface air temperature), accurately estimating the atmosphere variables through satellite remote sensing remains a significant challenge. This article introduces a deep learning transformer-based neural network (TaNet) for near-surface air temperature estimation. TaNet automatically extracts information from imageries captured by China's new-generation geostationary meteorological satellite FengYun-4A and generates grid near-surface air temperature data in near real time. Extensive experiments conducted using the state-of-the-art operational reanalysis product ERA5 and meteorological station observations as benchmark standards demonstrate the effectiveness and superiority of TaNet. It achieves an impressive Pearson's correlation coefficient (CC) of 0.990 with ERA5 and 0.959 with station observations, outperforming the other products, such as CFSv2, CRA, and U-Net, on root mean square error (RMSE) and Pearson's CC metrics. TaNet reduces the RMSE of CFSv2, CRA, and U-Net by a margin of 10.551$\%$ (2.594 $^{\circ }$C versus 2.900 $^{\circ }$C), 2.261$\%$ (2.594 $^{\circ }$C versus 2.654 $^{\circ }$C), and 5.535$\%$ (2.594 $^{\circ }$C versus 2.746 $^{\circ }$C), respectively, using station observations as the benchmark.

As a critical aviation support sensor, radar altimeters face imminent challenges due to the interference caused by the frequency overlap with the developing 5G telecommunications networks. This paper addresses the critical challenge posed by 5G interference with radar altimeter signals, which is crucial for maintaining aviation safety. It introduces a novel machine learning (ML) framework to preserve altimeter accuracy in the presence of 5G signals. This framework first classifies signals into pure or interfered categories, then applies regression models to predict altitudes when interference is detected, effectively quantifying and mitigating the interference impact. Distinguished by using real 5G signals from a Norman, Oklahoma base station, the approach offers a realistic evaluation and demonstrates the ML framework’s effectiveness in real-world conditions. Using various advanced ML models, the methodology showcases how critical aviation instruments can be safeguarded against the challenges posed by emerging telecommunications technologies, ensuring air travel safety and efficiency.

Compressed sensing theory is widely used to accurately reconstruct the original signal from a small number of random observations, i.e., obtain high-dimensional information from low-dimensional information. This feature has shown effectiveness in image super-resolution. In this paper, based on the compressed sensing theory, the ARSR (Adaptive Regularization term Super Resolution) algorithm is proposed to achieve super-resolution reconstruction of images. The algorithm models on the basis of exploiting the image local sparsity prior. On the one hand, by using adaptive regularization coefficients to weight the elements in the norm, we obtain more accurate sparse representation coefficients. On the other hand, we also add an adaptive regularization term behind the optimization model, which is able to take the correlation of the image into account as well. By generating a suitable coefficient, it can adaptively reconcile the relationship between sparsity and correlation. In addition, we derived an approximation to solve the model iteratively using the alternating direction method of multipliers (ADMM). In this paper, the sparse transform domain is trained by the K-SVD algorithm and the accuracy evaluation indexes of the experiments use peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). Compared with existing classical sparse representation-based image super-resolution algorithms, our ARSR algorithm obtains the highest PSNR and SSIM values in different regions of the image with better subjective visual effects.

Spectral clustering, a powerful algorithm in the field of AI, holds a significant role despite its inherent high time complexity. For data owners grappling with limitations such as small datasets and restricted computational resources, harnessing the computational capabilities of cloud computing and aggregating data from multiple sources can yield precise spectral clustering results. However, explicit data uploading to cloud servers poses privacy risks. In response to this challenge, we explore the outsourcing dilemma of spectral clustering in a cloud and multi-user environment and propose a quantum-secure and efficient solution. Specifically, by employing the CKKS homomorphic encryption algorithm within a dual non-collusive server model, we formulate a comprehensive and multi-user spectral clustering outsourcing scheme. Our approach addresses privacy concerns by introducing secure computation protocols for $L_{2}$ norm, exponential function, and negative half power function. We elaborate on efficient computational algorithms for each stage of spectral clustering, ensuring accurate clustering outcomes without compromising dataset privacy. Moreover, in our scheme, users only need to upload their encrypted dataset without requiring direct interaction with each other or the cloud server until obtaining clustering results. Finally, we argue the IND- $\mathcal {CPA}$ security of our design and substantiate its accuracy and efficiency through theoretical comparison analysis and experimental evaluations.

Autoencoders (AEs) have gained widespread application in the field of hyperspectral anomaly detection, largely due to their notable effectiveness in efficiently reconstructing backgrounds within hyperspectral images (HSIs). However, the absence of prior knowledge and constraints imposed by spectral information capacity hinder the accuracy of anomaly detection by allowing AEs to reconstruct both anomalous targets and backgrounds simultaneously. To address this limitation, a spatial–spectral cross-guided masked autoencoder (SSCMAE) has been proposed. The guided mask is generated based on the spectral difference between the anomaly and the background. This mask effectively suppresses the reconstruction of anomalous targets while enhancing the accuracy of background reconstruction. Moreover, a dual-branch structure operates, encompassing spatial and spectral dimensions, effectively capturing the inherent three-dimensional characteristics present in HSIs. Ingeniously designed cross-connection layers within the architecture enhance the spatial and spectral branches' capability of extracting internal spatial and spectral features of images. In order to capture a more comprehensive range of background features, a lightweight three-dimensional convolutional autoencoder is introduced. This addresses the issue of local feature loss during background reconstruction and overcomes the limitations that visual transformers face when learning local image structures. The proposed method has been systematically compared against several advanced methods on six real-world datasets. The results explicitly demonstrate the efficacy and superior performance of the presented SSCMAE approach.

The spatio-temporal action detection task requires the output of the temporal and spatial positions as well as the action category of the target action instances in the form of action tubes. However, the current definition of video-level metrics in spatio-temporal action detection tasks is not sufficiently clear and unified to fully describe the ability of network models to perform spatio-temporal detection. Furthermore, existing tube linking methods are not only heavily dependent on the quality of the detection stage but also lack reliable linking criteria, resulting in poor tube linking performance. To address these issues, this study proposes a hierarchical linking method based on multiple clues. This method first dynamically utilizes various correlation clues at two levels, including appearance features, spatial overlap, motion prediction, category scores, tube length, and tube confidence status, to reduce the negative impact of unreliable information on the correlation. Then, it employs inter-class correlation to handle the mutual influence between different categories, followed by joint probability data association to address the mutual influence between correlated objects, ultimately achieving robust and accurate online linking of action tubes. The method is experimentally compared with other correlation methods on the untrimmed UCF24 and MultiSports datasets, demonstrating state-of-the-art tube link performance. We also conducted ablation experiments to explore the impact of the different modules and stages in the proposed tube-linking method.

As a tool for near-earth remote sensing, unmanned aerial vehicle (UAV) can be used to acquire images and data of the earth's surface. This provides a powerful support for Earth observation and resource management. Object tracking in UAV videos has been a topic of much interest in recent years. A large number of algorithms have been proposed. Among these algorithms, deep learning has achieved a high accuracy rate. However, it is difficult to carry hardware devices for UAV, which makes it difficult to be practically applied. The correlation filter does not require a graphic processing unit to accelerate the computation, but it uses only manual features, which makes it difficult to achieve satisfactory performance. In order to solve the above problems, we proposed adaptive spatial regularization correlation filters, called DTSRT. Specifically, we first introduce deep features in the correlation filter instead of the original manual features, which can greatly improve the discriminative ability of the model. At the same time, in order to prevent affecting the real-time performance of the algorithm, we use histogram of oriented gradients features to determine the target scale and deep features to determine the target location. In addition, considering the large inter-frame distance between targets in UAV videos, we use a saliency detection method to dynamically generate spatially constrained templates. The proposed DTSRT outperforms other state-of-the-art algorithms with area under curve of 0.481, 0.431, and 0.474 on UAV123@10FPS, UAVDT, and DTB70 datasets, respectively.

Remote sensing change detection (CD) networks have been increasingly powerful with the application of convolutional neural networks (CNNs) and transformers. The CNN-based CD method with a CNN backbone has been widely used and plays an significant role. In complex spatial relationships within remote sensing images, CNNs may face limitations due to the restricted receptive field, making it challenging to handle intricate pixel relationships effectively. Therefore, to address this limitation of CNNs, we introduce vision graph neural network (ViG) to tackle the constrained receptive field issue. In addition, we propose a backbone network named Congraph, which integrates convolution and graph interaction. Congraph simultaneously leverages local information from CNNs and global information from GNNs, enabling more comprehensive feature extraction for more accurate change detection. Furthermore, we introduce a multiscale mixed attention (MMA) module to make the model focus on different scale feature information. MMA replaces small-scale features in the multilayer encoder with self-attention to capture global feature information within small-scale features. Finally, we feed bitemporal features into a transformer module to obtain feature difference information and generate the ultimate feature difference map. Through extensive experiments on the LEVIR-CD, WHU-CD, and GZ-CD datasets, our method demonstrates more significant performance advantages compared to the current state-of-the-art change detection methods.

The human body is a typical non-rigid object, and its 3D reconstruction is a classic problem in the field of computer vision. Due to the inherent complexity and dynamism of the human body, it is not suitable for existing non-rigid 3D motion reconstruction algorithms that assume that the number of shape bases of non-rigid bodies is known. The number of shape bases is very important for 3D reconstruction methods. If the number of shape bases is estimated incorrectly in contour deformation estimation, it can lead to unreliable or even complete failure of the reconstruction algorithm. Therefore, this paper designs a 3D non-rigid motion human object reconstruction algorithm based on reliable estimation of contour deformation degree. This algorithm leverage Scale Invariant Feature Transform (SIFT) to obtain non-rigid moving human target features. Firstly, the contour appearance model of the moving human sequence is used to extract the contour feature sequence, which is preprocessed based on the contour appearance depth feature; Furthermore, the deformation degree of the contour is reconstructed and the calculation process of the number of shape bases was optimized, which is no longer simply defined as known. This method optimizes and solves the problem of missing data, improves the reliability of estimating the degree of contour deformation, and completes target reconstruction. The experimental results show that the three-dimensional reconstruction algorithm can accurately reconstruct the changes in the movements of athletes’ shots; The accuracy of 3D reconstruction can reach 95.98%; Moreover, PSNR, SSIM, and MSE indexes performed well with smaller fluctuation range, and the distribution of three-dimensional reconstructed scattering points is very close to the three-dimensional position distribution of real scattering points, and the three-dimensional reconstruction effect is good with strong reliability.

In typical traffic scenarios such as non-signalized roads or shared spaces where vehicles and pedestrians interact without clear separations. The interaction distance between objects is usually shorter due to the simultaneous motion of road users. Pedestrian-crossing scenarios in these areas make the scenario complex due to the unpredictability of the pedestrians intention and the need to balance between safety, comfort, and time consumption. To address this collision avoidance(CA) problem, a novel strategy using Social Force Model (SFM)-based adaptive parameters was proposed. The interaction system between the ego vehicle and the pedestrian was simplified as a Markov process to adopt the SFM-based dynamic model, and the validity of this simplification was demonstrated using real-world driving data. Based on the current state of the interaction system that consists of vehicle and pedestrian, this research adopted the optimal parameters that were generated by particle swarm optimization (PSO) to generate optimal parameters for the SFM-based vehicle dynamic model, which helps the vehicle avoid pedestrians with random motion. The proposed method was validated through bench testing, and the results showed that the proposed method balanced the safety, comfort, and time consumption requirements during the CA process in the studied scenario.

Bicycles are frequently involved in traffic collisions with vehicles, particularly when sudden changes in direction occur. This paper presents a robust risk-predictive braking policy to ensure collision avoidance in all possible crossing behaviors of a bicycle. The policy controls the vehicle to follow an upper limit of the safe speed before the bicycle changes direction, ensuring that the vehicle can stop in time by the advanced emergency braking system before a collision occurs in any situation. The upper limit of the safe speed is the solution of an intractable robust optimization problem. Therefore, a scenario approach is adapted to develop a tractable approximate problem for the original robust optimization problem. The feasibility and optimality of the problem reduction are theoretically proved. A bisection method-based fast algorithm is designed to solve the approximate problem of the original robust optimization problem, making it applicable in practical scenarios. The convergence of the algorithm is also proven. The effectiveness of the proposed method is validated through hardware-in-the-loop simulations using CarMaker.

In this work, a metasurface-based, 1-bit electronically reconfigurable reflectarray antenna (RRA) is proposed for dual-polarization beam-steering applications. The proposed RRA consists of $16\times16$ Jerusalem cross-shaped unit cells loaded with PIN diodes, which are used as active and reconfigurable devices to generate a reflection phase difference of 180° between their on- and off-states. An optimal focal length to diameter ratio $(f/D)$ for the feed antenna is determined to maximize its radiation coverage on the RRA. Microcontroller-based shift registers are used to resolve the complexity of the individual DC bias networks for the PIN diodes. The proposed RRA is designed, fabricated, and measured to evaluate its beam-steering performance. The measurement results show that the proposed RRA can achieve beam-steering of ±60° for x- and y-polarizations in both azimuth and elevation planes. Moreover, the proposed RRA is further verified by full-wave simulations to evaluate its additional functions, such as the generation of sum and difference beams for monopulse radar systems. The proposed RRA is anticipated to have significant potential in areas such as wireless communications, imaging, and radar systems, and is a particularly suitable candidate for reconfigurable intelligent surfaces (RISs) due to its adaptability and capability of dynamically generating different radiation patterns.

To solve the problem of industrial control network (ICN) security situation prediction, this paper proposes a security situation prediction model for ICN based on evidential reasoning (ER) and belief rule base (BRB). First, this paper analyzes multiple factors influencing the security situation of the ICN, establishes a framework for security situation assessment, employs the ER algorithm for attribute fusion, and derives the security situation value of the ICN. Second, using historical data combined with expert knowledge, a security situation prediction model for ICN based on the BRB is constructed. Additionally, an extended projection covariance matrix adaptive evolution strategy (EP-CMA-ES) optimization algorithm is proposed, which is employed to optimize the parameters of the prediction model. The model not only comprehensively uses qualitative knowledge and quantitative data, but also integrates more uncertain information, and the reasoning process is interpretable. It also solves the subjectivity problem of expert knowledge, overcomes the problem of small amount of data caused by the difficulty of collecting industrial control safety data, and improves the accuracy of model prediction. Finally, prediction experiments were conducted on industrial datasets, confirming the feasibility and effectiveness of the security situation prediction model for ICN and the EP-CMA-ES optimization algorithm proposed in this paper.

The perception of voluntary respiratory consciousness is quite important in some situations, such as respiratory assistance and respiratory rehabilitation training, and the key signatures about voluntary respiration control may lie in the neural signals from brain manifested as electroencephalography (EEG). The present work aims to explore whether there exists correlation between voluntary respiration and scalp EEG. Evoke voluntary respiration of different intensities, while collecting EEG and respiration signal synchronously. Data from 11 participants were analyzed. Spectrum characteristics at low-frequency band were studied. Computation of EEG-respiration phase lock value (PLV) and EEG sample entropy were conducted as well. When breathing voluntarily, the 0–2 Hz band EEG power is significantly enhanced in frontal and right-parietal area. The distance between main peaks belonging to the two signals in 0–2 Hz spectrum graph tends to get smaller, while EEG-respiration PLV increases in frontal area. Besides, the sample entropy of EEG shows a trend of decreasing during voluntary respiration in both areas. There’s a strong correlation between voluntary respiration and scalp EEG. Significance: The discoveries will provide guidelines for developing a voluntary respiratory consciousness identifying method and make it possible to monitor people’s intention of respiration by noninvasive BCI.

Faults are a type of geological structure easily “reactivated” by underground mining, which destroys the internal movement and deformation laws of rocks and soil masses and causes anomalous surface sinking features. Therefore, studying the spatiotemporal evolution and deformation mechanism of mining subsidence in regions of fault occurrence is crucial for logical planning of the operating face and preventing and providing early warning of geological disasters. Considering the abnormal surface damage phenomenon in a town in the Hebei Province as the research background, this study used 46 Sentinel-1A radar images from January 3, 2019, to June 26, 2020, to investigate the spatiotemporal evolution of surface movement and deformation during the mining of the working face using the time-series InSAR deformation analysis method. Particularly, a four-threshold permanent scatterer point selection method was proposed and applied to small baseline subset InSAR monitoring. Compared with the monitoring data of 69 benchmarks in the same period, the mean absolute error and root mean square error between InSAR and leveling were 6.7 and 5.3 mm, respectively. These values indicated that the deformation of the mining area inverted by InSAR was reliable. Subsequently, the response mechanism of surface deformation in a region of fault occurrence was analyzed based on separation space theory using a rock beam mechanical model. The results revealed that the abnormal damage area had been deformed since September, 2019, and the abnormal deformation continued to develop with the advancement of the working face. The termination position of the abnormal deformation was approximately linear and relatively spatially independent of the conventional subsidence basin. Additionally, the time-series deformation characteristics of different regions within the influence range of mining were different, and the surface subsidence curve in the direct influence area conformed to the conventional law of subsidence. Surface subsidence in the indirect influence area experienced a process of acceleration and then slowed; the deformation mechanism was different from the conventional movement law of rock strata. The surface outside the affected area exhibited a small uplift, and its active time was consistent with the fault slip instability time. Furthermore, after the fault was disturbed, the rock mass in the fault zone slipped towards the goaf, and a separation space was generated at the fault plane and transferred to the loose layer, which promoted the abnormal expansion of the movement range of overburden and severe deformation at the outcrops. Simultaneously, the footwall of the fault was slightly uplifted owing to the lever principle. Our research results provide data-driven support for the study of abnormal damage laws and the disclosure of their causes. These findings are imperative for the development and improvement of the subsequent subsidence control theory.

Infrared small target detection is widely used in military and civil security fields. Most of the existing infrared small target detection algorithms focus on improving detection accuracy. However, the lightweight and generalization capabilities of the network have not improved significantly. To improve the network's generalization ability and lightweight while maintaining accuracy, a dense nested network based on position-aware dynamic parameter convolution kernel (Par-DPC DNNet) is proposed. First, a generation module for the dynamic parameter convolution kernel is proposed. The convolution parameters are generated dynamically based on the instance. This can realize the network's local dynamic change for unfixed small targets and effectively improve the network's generalization ability. Moreover, this can achieve good detection performance on the basis of network depth reduction. Second, in order to apply the dynamic parameter convolution kernel in the network on a large scale, inspired by the patch fragment mode in the VIT structure, a new downsampling method is proposed. The dynamic multiples of downsampling are realized by the feature map fragment combined with the convolution method without information loss. Finally, the dynamic parameter convolution kernel and new downsampling are integrated into the dense nested network, and a new nonlinear feature extraction method is adopted. It improves network accuracy while maintaining network generalization and structure optimization. The effectiveness of the proposed method is verified using two datasets. The experimental results demonstrate that the detection performance of Par-DPC DNNet is superior to the existing state-of-the-art methods even if the number of internal layers of Resblock is reduced to two.

Existing image processing methods usually divide image denoising and image fusion into two directions for research, and even the best current image denoising methods such as DnCNN can cause information loss during image processing, and the image fusion method mainly considers the fusion between multiple source images to complement the image information, but does not take into account the degradation of the fusion quality due to the noise in the source images. Therefore, aiming at the problem that various existing image denoising methods cannot reduce noise efficiently in complex noise situations while performing multi-source image fusion, preserving texture details, highlighting edge contour structures, and enriching image energy, a method of reconstructing remote sensing images by simplified adaptive dual-channel PCNN (Dual-PCNN) fusion in the NSCT transform domain is proposed to unify image noise reduction and image fusion under the same framework to obtain a noisy image information reconstruction method, which completes the complementary advantages between two image processing methods. Firstly, the impulse noise in the image is removed using IMFLED filtering, and then the Gaussian noise is processed by DnCNN and FFDNet respectively, and the results are imported into simplified parametric adaptive dual-channel PCNN to fuse the preprocessed images respectively. Then the DnCNN image and the preprocessed image are decomposed by NSCT to obtain each low-frequency sub-band and high-frequency sub-band. The low-frequency sub-band is stimulated with detail using the guided filter, and the high-frequency sub-band is enhanced by separating the detail layer through the guided filter to obtain the energy-detail-enhanced high-frequency sub-band. Finally, the corresponding sub-bands are fused into the simplified parametric adaptive dual-channel PCNN respectively, and the fused sub-band coefficients are reconstructed by NSCT inversion to obtain the final reconstructed remote sensing image. The experiments on grayscale images and remote sensing images show that this method achieves excellent results in both visual subjective and quantitative index evaluation, and the reconstructed images perform well in texture details, contour structure, and energy enrichment. After the information reconstruction of remote sensing images, the quality and resolution of remote sensing images are effectively improved, so that terrain information, landform features, and structural features can be extracted more accurately, which are widely used in remote sensing and geographic information fields such as landform analysis, agricultural monitoring, building inspection and environmental protection.

Supervision of Livelihood Issues (SLI) refers to the supervision and participation of the general public in government actions and public affairs, with a particular focus on issues closely related to people’s lives. However, current regulatory systems often suffers from information imbalances, low authenticity, and credibility. To address these challenges, this study proposes a blockchain-based SLI framework that leverages the decentralized, tamper-proof, traceable, and transparent nature of blockchain technology to provide a reliable platform for SLI. Furthermore, we designed an information data structure on the blockchain and presented the information flow process. We developed smart contracts that utilize automation capability to automatically aggregate, analyze, and publish warnings based on on-chain data, enabling the evaluation of relevant departments’ performance. Moreover, in the context of SLI applications, we propose an optimized consensus algorithm called the SLI-PBFT. This algorithm is based on the PBFT consensus protocol and incorporates dynamic scalability mechanisms, scoring models, and simplified consistency protocols to enhance the consensus efficiency of SLI applications. Finally, the experiments were conducted. We developed an SLI prototype system and compared and analyzed the performance of SLI-PBFT with other PBFT-based consensus algorithms within the SLI prototype system. The experimental results demonstrated the feasibility of the SLI framework. Moreover, SLI-PBFT exhibits significant improvements in throughput and reduced latency compared to other algorithms in the context of SLI applications. In addition, they possess strong fault tolerances and security capabilities. This study provides theoretical and practical guidance for the SLI in real-world scenarios.

A method for estimating atmospheric relative humidity using wind profiler radar and a “cascaded” machine learning algorithm is introduced. Unlike existing methods in the literature, the proposed approach uses only I/Q or moment data from the profiler radar to generate an intermediate pressure profile, which serves as training data for humidity estimations without requiring temperature as an input feature. The study examines the potential of various machine learning algorithms and evaluates their performance using field data collected by the Hong Kong Observatory between January and June 2021. Importantly, this is the first time a cascading machine-learning solution has been successfully applied to the humidity estimation problem, resulting in a simplified model with reduced complexity and fewer required features.

The purpose of object tracking is to locate a given target in image sequence, such as people and vehicles. In recent years, with the development of unmanned aerial vehicle (UAV) technology, object tracking in UAV videos has engaged many scholars. It has been widely used in traffic control, water quality inspection, wildlife census, and other fields. However, low resolution, scale change, occlusion, and other challenges have been restricting the development of the tracker. To solve the aforementioned problems, we put forward multifeature correlation filters with saliency proposals. First, we use histogram of oriented gradient features, gray (I) features, and color names features to heighten the representation information of the target, so that our algorithm can accurately locate small targets. Then, we introduce saliency proposals to reposition the occluded target. Finally, we use dynamic update weights instead of the fixed update weights to mitigate the adverse effects caused by template degradation. Experiments demonstrate that our tracker has achieved satisfactory tracking accuracy and AUC scores have reached 0.462, 0.417, and 0.425 on UAV123@10FPS, UAV20 L, and UAVDT datasets, respectively.

During the production and assembly process of televisions, the efficiency of assembly is limited by manual assembly methods. Conventional image matching algorithms such as SIFT and ORB fail to meet the real-time accuracy requirements of intelligent assembly processes. To address this issue, a novel algorithm called Adaptive Threshold oFAST and rBRIEF-BRISK (AFK) is proposed. This algorithm utilizes the Hamming distance as a similarity measure and integrates AToFAST and rBRIEF-BRISK fusion feature descriptors. Through a comparison of 12 sets of image matching scenarios, it is demonstrated that the AFK algorithm achieves a matching speed 4.5 times faster than the SIFT algorithm, with an image matching accuracy of up to 90%. Additionally, this paper presents the GLMLBP-TOP (Global-Local and Michelson Contrast LBP-Three Orthogonal Planes) dynamic texture recognition method. Through a comparison of 21 sets experiments, it is shown that the comprehensive classification accuracy of the GLMLBP-TOP method is approximately 94.5%, improving the performance of VLBP (87.71%) and LBP-TOP (91.1%) by 6.79% and 3.4% respectively. Real-world application results demonstrate that this method meets the requirements of intelligent assembly processes and has been successfully applied to the LB intelligent assembly process of a certain company’s televisions.

The network security status assessment (NSSA) method can evaluate the network security status of cloud manufacturing systems (CMSs), which is of great significance to reduce the network security risk and loss of CMSs. At present, the NSSA of CMSs suitable for semiquantitative and uncertain information conditions is commonly used, which has certain limitations and low accuracy. This paper proposes an NSSA method for CMSs based on semiquantitative information. First, through the detailed analysis of the influencing factors of the network security status of the CMSs, the security evaluation indicators are selected, and the evaluation framework containing multilevel indicators is established by combining semiquantitative information. Second, the evaluation level is established according to the data distribution and the properties of the indicators. Third, a process of data fusion reasoning based on an evidential reasoning algorithm is designed, and a strong reference fusion case is given. Finally, the effectiveness of the proposed security state assessment algorithm in the NSSA of CMSs is verified and analyzed by simulation experiments. The experimental results show that the proposed method can make full use of semiquantitative information and uncertain information to evaluate the network security status of CMSs, and the evaluation results can reflect the actual security status of CMSs.

The establishment and improvement of an innovation ecosystem have become important issues in building an innovative powerhouse. Built on the background of innovation ecosystem, from the perspective of cooperative interaction and dynamics, this study introduces value co-creation and dynamic capabilities, and analyzes how the proximity between enterprises and scientific research organizations and between enterprises affects enterprise innovation performance. The results demonstrate that, on the one hand, the proximity between enterprises and scientific research organizations and between enterprises can directly have an inverted U-shaped effect on enterprise innovation performance. On the other hand, it can indirectly have an inverted U-shaped effect on enterprise innovation performance through the partial mediating effect of value co-creation. Dynamic capabilities play a major role in regulating the relationship between proximity and enterprise innovation performance, as well as value co-creation and enterprise innovation performance. The research results expand the research in the field of innovation ecosystems, reveal the marginal effect of value co-creation on enterprise innovation performance, and help enterprises in the system to clarify how to improve innovation performance through proximity strategies and value co-creation activities with scientific institutions and other enterprises.

Network intrusion detection is an important technology in national cyberspace security strategy and has become a research hotspot in various cyberspace security issues in recent years. The development of effective and efficient intelligent network intrusion detection methods using advanced machine learning algorithms is of great importance for defending against various network intrusions in complex network environments. In this study, a network intrusion detection method based on decision tree twin support vector machine and hierarchical clustering, named HC-DTTWSVM, is proposed, which can effectively detect different categories of network intrusion. First, the hierarchical clustering algorithm is applied to construct the decision tree for network traffic data, where the bottom-up merging approach is used to maximize the separation of the upper nodes of the decision tree, which reduces the error accumulation in the construction of the decision tree. Then, twin support vector machines are embedded in the constructed decision tree to implement the network intrusion detection model, which can effectively detect the network intrusion category in a top-down manner. The detection performance of the proposed HC-DTTWSVM method is evaluated on NSL-KDD and UNSW-NB15 intrusion detection benchmark datasets. Experimental results show that HC-DTTWSVM can effectively detect different categories of network intrusion and achieves comparable detection performance compared to some of the recently proposed network intrusion detection methods.

In this article, a method of extending airborne weather radar modeling to incorporate high-ice-water-content (HIWC) conditions has been developed. A novel aspect is incorporating flight test measurement data, including forward-looking radar measurements and in situ microphysics probes data, into the model and part of the evaluations of modeling. The simulation models assume a dual-polarized, airborne forward-looking radar, while for single-polarized operations, the index-of-dispersion is included as a helpful indicator for HIWC detection. The radar system simulation models are useful for design evaluations for the next generation of airborne aviation hazard monitoring and incorporate HIWC hazard detection algorithms. Example applications of the simulator, such as hazard detection based on the simulated HIWC flight encounter scenarios based on specific numeric weather prediction (NWP) model outputs, are discussed.

This paper describes the design of a bionic soft exoskeleton and demonstrates its feasibility for assisting the expectoration function rehabilitation of patients with spinal cord injury (SCI). Methods: A human–robot coupling respiratory mechanic model is established to mimic human cough, and a synergic inspire–expire assistance strategy is proposed to maximize the peak expiratory flow (PEF), the key metric for promoting cough intensity. The negative pressure module of the exoskeleton is a soft “iron lung” using layer-jamming actuation. It assists inspiration by increasing insufflation to mimic diaphragm and intercostal muscle contraction. The positive pressure module exploits soft origami actuators for assistive expiration; it pressures human abdomen and bionically “pushes” the diaphragm upward. Results: The maximum increase in PEF ratios for mannequins, healthy participants, and patients with SCI with robotic assistance were 57.67%, 278.10%, and 124.47%, respectively. The soft exoskeleton assisted one tetraplegic SCI patient to cough up phlegm successfully. Conclusion: The experimental results suggest that the proposed soft exoskeleton is promising for assisting the expectoration ability of SCI patients in everyday life scenarios. Significance: The proposed soft exoskeleton is promising for advancing the application field of rehabilitation exoskeletons from motor functions to respiratory functions.

With the advent of big data era and the rapid improvement of raw data scale, feature selection, as the basis and critical technologies for data mining, plays an increasingly important role. However, most studies on feature selection methods, mainly directed to treat the single feature or overall feature subset, while the influence of the correlation and redundancy of features in the feature subset on the classification results is ignored. In this paper, a hybrid feature selection method based on feature subsets generated by factor analysis (FAFS_HFS) is proposed. Firstly, this method generates feature subsets from the maximum load (maximum explanatory power) of each feature through factor analysis. Then, minimal redundancy and maximal relevance (mRMR) and sequential forward selection (SFS) are used to remove the redundancy of each feature subset. Finally, fisher score based on feature subset (FSF-score) is utilized to evaluate and obtain the optimal feature subsets. Experiments are conducted on 14 datasets, the results show that FAFS_HFS method has higher classification accuracy and lower dimension on almost all datasets, especially in high-dimensional datasets, and it has competitive efficiency and classification performance compared with other methods.

A systematic approach to synthesize the flat-topped beam (FTB) with arbitrary beam directions for uniform linear arrays (ULAs) is proposed, illustrated, and demonstrated. Based on the correspondence between the far-field pattern of ULAs and the response function of finite impulse response (FIR) filters, both the broadside and end-fire FTBs can be synthesized using window method, whereas the oblique FTB synthesis is failed due to the appearance of grating lobes. A mathematical analyzation indicates that two grating lobes are conjugate to each other. Thus, a linear phase method (LPM) is proposed to apply progressive excitation phases to each individual array element, which can eliminate the grating lobes by manipulating the two conjugate beams to overlap with each other. The proposed LPM is generalized to the synthesis of broadside, oblique and unidirectional end-fire FTBs, and the corresponding inter-element spacing constraints are addressed. The proposed approach is experimentally verified using a 16-element substrate-integrated waveguide slot antenna array. Comparing with the conventional Fourier series method and Woodward-Lawson method, the proposed approach has advantages of both smaller ripples in the main beam and controllable side-lobe levels. The proposed approach is promising for offering efficient beamforming functionalities for wireless communications, imaging, and radar applications.

Interconnect development for the new technology node requires coordinated efforts of multiple module teams working to co-optimize patterning and metallization solutions to meet performance and yield, and reliability targets. This paper presents the results of interconnect patterning optimization to improve overlay and process window control using a novel methodology, Design-for-Inspection™ (DFI). Using design-assisted voltage contrast measurement techniques, the method enables in-line test and monitoring of process induced overlay and CD variation of back-end-of line (BEOL) features built with litho-etch-litho-etch (LELE) patterning. While only some of the features of multi-color patterning scheme are chosen to be aligned directly, other combination of metal line and via colors may have uncontrolled misalignment risking open or short failures. The paper shows how the complete metrology coverage of multi-color combinations between dual patterned via and dual patterned metal lines helps driving the improvement of overlay and process margins in 14nm technology. The optimized process margin for via opens enables higher yields and better reliability.

At present, visual and parametric designs have been widely used in architectural design, urban planning and landscape design for design inspiration. Highly integrated auxiliary design software has become more prevalent and greatly changed the traditional way of working with architectural design and urban planning with parameterised operations, convenient editing methods and efficient workflows. However, these methods for layout optimisation have more or fewer problems, such as humdrum forms, lack of aesthetic principles, and difficulty adapting to the layout optimisation of complex architectural environments, and usually presented obstacles to the users’ understanding of hierarchical relationships in complex environments. This paper presents a novel 3D scene layout optimisation method that combines scene view information with an interactive genetic algorithm to optimise the virtual environment’s layout. We apply the visual perception information of 3D building models to improve the scene’s layout and integrate the evaluation of visual salient features with other factors to realise the intelligent scene’s layout based on optimal 3D viewpoints and a genetic algorithm. The experimental results show that our method can efficiently optimise the layout of the 3D building environment, and it has innovation and practical application value in the fields of architectural visualisation, 3D animation, landscape planning and game design.

An interesting application of the link prediction technique is detecting the potential new links in collaboration networks. In this study, we construct collaboration networks based on the co-authorship information of the papers published in 43 statistical journals from 2001 to 2018. We construct training and testing networks according to the timestamps of the papers and construct a classification dataset for link prediction. We calculate 20 similarity indices based on the training network to perform link prediction. Additionally, we consider nodal attributes (institutes and research interests) to develop two novel predictors called the same institute (SIN) and keywords match count (KMC). Several machine-learning classifiers including support vector machine, XGBoost and random forest are implemented to combine all predictors. After incorporating SIN and KMC, we observe that the area under the receiver operating characteristic curve values of all classifiers improved, indicating that SIN and KMC can significantly improve classification accuracy. Finally, we provide collaborative recommendations for researchers based on the proposed model.

An adaptive on-line approximator-based finite-time trajectory tracking scheme is proposed to solve the problem of trajectory tracking of underactuated surface vessel affected by dynamic uncertainties and external disturbances. In this scheme, underactuated transformation is performed by using kinematic virtual control law transformation and bounded constraint. By designing adaptive on-line approximator to approach the upper bound of uncertainty and unknown disturbance, the problem of parameter uncertainty and external disturbance are solved. It is proven by Lyapunov theory that the error signals in the system can converge quickly to the stable region in finite time. Finally, by comparing the simulation results, it is verified that the proposed control scheme can make the underactuated ship track the desired trajectory in finite time. Compared to the traditional control method, the convergence rate of the system error is faster, and it exhibits strong robustness in the face of unknown external disturbances. This has a certain reference value for practical engineering applications.

In response to the imbalance between power generation and demand, demand response (DR) projects are vigorously promoted. However, customers' DR behaviors are still difficult to be simulated accurately and objectively. To tackle this challenge, we propose a new DR behavioral learning method based on a generative adversary network to learn customers' DR habits. The proposed method is also extended to maximize the economic revenues of generated DR policies on the premise of obeying customers' DR habits, which is hard to be realized simultaneously by existing model-based methods and traditional learning-based methods. To further consider customers' time-varying DR patterns and trace the changes dynamically, we de-fine customers' DR participation positivity as an indicator of their DR pattern and propose a condition regulation approach improving the natural generative adversary framework to generate DR policies conforming to customers' current DR patterns. The proposed method is applied to hourly electricity price optimization to reduce the fluctuation of system aggregate loads. An online parameter updating method is also utilized to train the proposed behavioral learning model in continuous DR simulations during electricity price optimization. Finally, numerical simulations are conducted to verify the effectiveness and superiority of the proposed method.

To solve the problem of low message delivery rate and high network resource consumption when forwarding messages in opportunistic networks, an opportunistic routing algorithm based on trust relationships for wireless mesh networks is proposed. Firstly, the wireless mesh network is analyzed and the opportunistic routing model is constructed; By analyzing the security mechanism and security threat of communication entities, then measuring the trust degree of links and nodes, establishing the trust relationship between nodes, and defining and quantifying a new security measurement method based on the trust model; Finally, according to the security measurement method defined by the model, select the node with high trust value to participate in the message forwarding process. At the same time, give priority to the node with greater trust with the destination node as the relay node, and allocate the message copy according to the trust degree to make the message pass along the direction of increasing trust, to complete the design of opportunistic routing algorithm in wireless mesh networks. Experimental results show that the routing algorithm can effectively improve the message delivery rate, up to about 95%, and reduce the consumption of network resources.

The well-known compressed sensing reconstruction ( $\mathcal {CSR}$ ) uses the sparse characteristics of the signal to obtain discrete samples with the compression (i.e. measurement) algorithm, and then perfectly reconstructs the signal through the reconstruction algorithm. Benefiting from the storage savings, the $\mathcal {CSR}$ has been widely used in the field of large-scale image processing. However, the reconstruction process is computationally overloaded for resource-constrained clients. Therefore, designing a cloud-aided $\mathcal {CSR}$ algorithm becomes a hot topic. In this paper, we investigate the existing secure $\mathcal {CSR}$ algorithms within a cloud environment and propose a new privacy-enhanced and verifiable $\mathcal {CSR}$ outsourcing algorithm for online medical image processing services. Compared with previous work, our new design can efficiently achieve more extensive security. Precisely, (1) our algorithm realizes the privacy preservation of the original image, as well as the input/output information of the reconstruction process under the chosen-plaintext attack, (2) our design is based on a malicious cloud server model and can verify the correctness of the cloud returned result with a probability of approximating 1, and (3) our algorithm is highly efficient and can make the local client achieve decent computational savings. The main technique of our design is a combination of linear transformation, permutation and restricted random padding which is concise and high-efficiency. We analyze the above claims with rigorous theoretical arguments and comprehensive experimental analysis.

To improve the performance of anti-jamming communication in dynamic and adversarial jamming environment, an improved anti-jamming method is proposed based on deep reinforcement learning and feature engineering. Different from the existing studies that use computer vision of deep learning based on the infinite state of spectrum waterfall, the proposed method relays on analyzing spectrum differences between adjacent time slots which contains information and features of jamming patterns. First, anti-jamming strategy is trained by countering the jammer which carries out a random jamming patterns switching strategy. Second, an improved state space is introduced by containing historical spectrum of communication and jamming signal between adjacent time slots, which can help an anti-jamming agent effectively extract the features of jamming patterns to reduce computational complexity. In addition, an improved reward function based on channel switch cost is improved for considering propagation characteristics which may cause communication performance lost. Taking advantage of both feature engineering and deep reinforcement learning, an improved anti-jamming method is proposed to improve reliable anti-jamming performance. Compared with the traditional CNN-based deep reinforcement learning anti-jamming method, simulation results show that the improved method can obtain better performance and lower computational complexity.

Social identity theory and self-classification theory provide theoretical supports for explaining the formation and action of the technological divisive faultlines. The purpose of this study is to explore the effects of technological divisive faultlines on a clique’s innovation performance and the moderating effects of geographical proximity and closeness centrality. The data of 119 firm alliance and 351 firms in cliques are collected in China’s biomedical industry from 2010-2017. Then, we perform the negative binomial regression by STATA software with these data as samples. The results show that there is an inverted U-shaped relationship between the technological divisive faultlines and clique’s innovation performance; geographical proximity has no moderating effect on the inverted U-shaped relationship; closeness centrality has a significant negative moderating effect on the inverted U-shaped relationship between them. Besides, it suggests that the clique’s innovation performance under high closeness centrality is higher than that under low closeness centrality, which provides a new perspective for the previous research that forming cliques could break the relational redundancy.

This article presents development of CMOS system-on-chip spectrometer processor for Spaceborne Earth Science, Planetary Science and Astronomy. The developed chip is intended for use in rotational emission spectroscopy observations from microwave to sub-millimeter wavelengths performed from Earth orbiting or deep space exploration spacecraft. The article first provides an overview of rotational spectroscopy for space sciences and highlights some of the important results achieved in Earth science, planetary science, and astronomy. The article then discusses the need for smaller and more compact spectrometer instruments across these three space science areas, which motivates this work. Then key considerations in development of spectrometer processors for space science are discussed including the significance including bandwidth, quantization efficiency and spectral channel shapes. The article then highlights detailed design of the developed spectrometer processor focusing on the many calibration techniques included to allow the chip to adapt to operations in space environments where large temperature ranges and high radiation levels are encountered. Presented measurements demonstrate the spectrometer processor operates up to 12.8 GS/s to provide up to 6.4 GHz of processing bandwidth with a frequency resolution of 8192 spectral channels and an analog resolution of 4 bits while consuming only 3960 mW of total DC power.

A Ka-band 1024-channel passive millimeter wave (PMMW) imager (BHU-1024) based on synthetic aperture interferometric radiometer (SAIR) technique has been developed by Beihang University for security screening. BHU-1024 uses linear phased array to obtain resolution in the horizontal direction and uses aperture synthesis to obtain resolution in the vertical direction. This imager is designed for detecting concealed weapons on the human body and operated under the near-field condition of the antenna array. Thus, the conventional direct Fourier imaging theory, which is based on the far-field approximation, can’t be applied any longer. In this paper, a novel near-field image reconstruction method based on the beamforming technique is proposed. In this method, we derive the near-field imaging formula by beamforming theory and demonstrate the feasibility of obtaining spatial brightness temperature distribution by beam focusing. In the design of the beamformer, we further optimize the weight vector to suppress the sidelobe levels of the synthesized beam. The results of simulation and measurement demonstrate that the proposed method is an advantageous, effective imaging method for near-field passive millimeter wave imaging. At present, the proposed method has been applied to the actual passive millimeter wave imaging system BHU-1024.

In the field of human-machine interaction, gesture recognition using sparse multichannel surface electromyography (sEMG) remains a challenge. Based on the Hilbert filling curve, a dual-view multi-scale convolutional neural network (DVMSCNN) is designed to enhance gesture recognition performance in this paper. The network consists of two parts. In the first part, sEMG is filled using Hilbert filling curve, and the obtained images in the time and electrode domain are used as inputs to the block. In the second part, the depth features learned by block are fused and classified by a “layer fusion” based view aggregation network. The evaluation of the architecture in the four databases of Ninapro-DB1, DB2, DB3 and DB4 shows that DVMSCNN is more than 7% more accurate than other state-of-the-art methods. When validated using a home-grown dataset, DVMSCNN was able to achieve a recognition rate of 0.8848.

Aiming at the problems of unstable motion and low tracking accuracy caused by complex external disturbance during the movement of the remote operated vehicle(ROV), the adaptive control method and sliding mode control method are combined to propose a ROV adaptive sliding mode motion controller(ASM controller). The sliding mode surface is designed by exponential reaching law and saturation function to achieve rapid convergence of the control system and eliminate high-frequency buffeting, combined with adaptive algorithm to improve the anti-disturbance ability of the system, and the Lyapunov stability criterion is used to verify the controller's stability under uncertain parameters and unknown external disturbances. Simulation experiments show that the designed adaptive sliding mode controller has good maneuverability and tracking performance.

Variational models with second order regularizers can efficiently overcome the problems of staircasing effects caused by first order models. However, different types of second order regularizers may lead to different properties of feature preserving in restored images. In this paper, we show two variational models with second order regularizers. The first one is bounded Hessian model with Jacobian of normals, which uses bounded Jacobian of image intensity normals as regularizer, it is an extension of classical bounded Hessian (BH) model. The second one is total generalized variation model with Jacobian of normals, which is an extension to total generalized variation (TGV) model by replacing gradients in TGV with normals. The common objective is to improve feature preserving, such as edge, contrast and smoothness preservation. Additionally, their Alternating Direction Method of Multipliers (ADMM) are designed by introducing some proper auxiliary variables, Lagrange multipliers and penalty parameters to decompose the original models into some simple minimization sub-problems to solve. Extensive comparisons demonstrate that the proposed models are superior to the classical models with Hessian and TGV regularizers, especially in edge and corner preservation, smoothness, contrast enhancement. Moreover, the proposed models can be also extended to image inpainting, deblurring, and image enhancement.

The increasing popularity of smart vehicles and vehicular ad hoc networks (VANETs) has reinforced the importance of Internet connectivity. However, existing solutions (e.g., those based on cooperative downloading via drive-thru Internet) have a number of limitations, partly due to the challenges in selecting reliable proxy vehicles in a fast moving driving situation, and balancing between flexible data downloading and achieving strong security. Therefore, in this paper we propose a secure, robust and flexible cooperative downloading scheme based on our reputation based selection mechanism and ordered signatures. Using the reputation based selection mechanism, only vehicles with the highest expected downloading capacities will be selected as proxy vehicles. This helps us to avoid selecting less reliable proxy vehicles. The flexible data downloading is achieved by dividing a file into small blocks so that a proxy vehicle can flexibly vary the file / data download, for example based on existing condition. Our scheme also achieves strong security, in the sense that it realizes traditional authentication, privacy preservation and message confidentiality, as well as a newly introduced security requirement (i.e., process authentication: authenticates the order of file blocks and the order of the assisting vehicles using ordered signatures).

Classification is one of the most classic problems in machine learning. Due to the global optimization ability, evolutionary computation (EC) techniques have been successfully applied to solve many problems and the evolutionary classification model is one of the methods used to solve classification problems. Recently, some evolutionary algorithms (EAs) such as the fireworks algorithm (FWA) and brain storm optimization (BSO) algorithm have been employed to implement the evolutionary classification model and achieved the desired results. This means that it is feasible to use EC techniques to solve the classification problem directly. However, the existing evolutionary classification model still has some disadvantages. The limited datasets used in the experiment make the experimental results not convincing enough, and more importantly, the structure of the evolutionary classification model is closely related to the dimension of datasets, which may lead to poor classification performance, especially on large-scale datasets. Therefore, this paper aims at improving the structure of the evolutionary classification model to improve classification performance. Feature selection is an effective method to deal with large datasets, firstly, we introduce the concept of feature selection and use the different feature subsets to construct the structure of the evolutionary classification model. Then, the BSO algorithm is employed to implement the evolutionary classification model to search the optimal structure by search for the optimal feature subset. Moreover, the optimal weight parameters corresponding to the different structures are also searched by the BSO algorithm while searching the optimal feature subset. For verification of the classification effectiveness of the proposed method, 11 different datasets are selected for experiments. The results show that it is feasible to optimize the structure of the evolutionary classification model by introducing feature selection. Moreover, the new method has better classification performance than the original method, especially on large-scale or high-dimensional datasets.

Test suite reduction is an effective way to reduce the cost of regression testing by identifying and removing redundant test cases from the original test suite. In this paper, we propose a novel cluster test suite reduction using an evolutionary multi-objective optimization algorithm. Specifically, we use a K-means algorithm to group similar test cases to the same cluster. Then the evolutionary algorithm is used to remove redundant test cases based on the clustering results, and optimization objects are represented as the coverage-related criteria, fault-related criteria and cost-related criteria. The experimental results involving eight subject programs show that the proposed method can outperform the other three state-of-the-arts with respect to both fault detection (4.61% -9.44%) and reduction ratio (4.10% -10.64%). Meanwhile, the experiments also prove that our method has a better performance of missing failure rate (0.049% -0.132%) and code coverage rate (3.34% -6.10%). Besides, the proposed method costs are found to be comparable to the other techniques.

In order to improve classification performance and operating efficiency of the hash algorithm, this paper proposes a novel hash algorithm that combines three-dimensional global features and local energy features. During the stage of three-dimensional features extraction, the image is firstly compressed by SVD decomposition to form a secondary image. Then the statistical features of the secondary image at the three-dimensional visual angle are extracted as the global features. Finally, the global feature hash is generated by using the relationship between the statistical features of the image layers from different three-dimensional visual angles. In the energy feature extraction stage, the luminance image is divided into blocks, and then the energy value of each image sub-block is obtained. The multi-directional energy change features are taken as the local features of the image. Subsequent experimental results prove the effectiveness of the proposed algorithm. The algorithm not only has good robustness to the conventional content-preserving operations, but also achieves a good balance between discrimination capability and robustness. In addition, compared with several state-of-the-art schemes, this algorithm has the best ROC curve, the shortest running time and the best local tamper detection ability.

Specific emitter identification (SEI) uses the unintentional modulation information carried by the emitter waveform, i.e., radio frequency fingerprints, to realize the matching identification of the received signal and its corresponding emitter. We propose an emitter identification scheme based on deep learning (DL). The received signal is subjected to time-varying filtered empirical mode decomposition (tvf-EMD). The obtained intrinsic mode functions (IMFs) are subjected to Hilbert transformation to obtain a 3D-Hilbert spectrum, using amplitude frequency aggregation characteristics obtained by projection to express the nonstationary information of the signal and calculate its bispectral diagonal slice as the second feature. A squeeze-and-excitation block (SEB) is introduced to a convolutional neural network (CNN) to focus on the effective area in the feature map, and support vector machine (SVM) is used to fuse the decision information of the two types of features. We collect four types of steady-state signals on a software-defined radio (SDR) sensor platform based on GNU Radio and universal software radio peripherals (USRP), and study the algorithm performance. The results show that our scheme has strong generalization, high recognition accuracy, and broad application prospects compared with other SEI methods.