最新刊期
卷 53 ， 期 3 ， 2025
- 摘要：As the cost of unmanned aerial vehicles (UAVs) decreases, they have attracted increasing research interest. UAVs are now widely applied in various fields, including agriculture, firefighting, surveying, aerial photography, and recreational applications. These applications require UAVs to perform autonomous flights with precise self-localization, typically relying heavily on global navigation satellite systems (GNSS). However, GNSS has multiple shortcomings related to long-distance radio communications, such as non-line-of-sight reception, multi-path effects, and spoofing. This has driven the development of new methods to supplement or replace satellite navigation. Vision-based UAV localization and navigation methods, utilizing onboard visual sensors for autonomous localization and navigation, have become crucial in addressing this issue. This review contributes to the field by systematically reviewing vision-based UAV localization and navigation technologies, providing a comprehensive summary of the current research landscape and developmental trends. First, it introduces vision-based UAV localization methods, which are categorized into image retrieval and feature matching approaches. The technical characteristics, applicable scenarios, relevant datasets, and evaluation metrics of these methods are analyzed in detail. Second, this review elaborates on vision-based UAV navigation methods, distinguishing between obstacle detection and avoidance techniques and path planning methods based on their functional objectives, while highlighting the strengths and limitations of existing technologies. Finally, this review further discusses the potential challenges faced by vision-based UAV localization and navigation methods, including the lack of publicly available datasets, the need for hardware acceleration, the complexity of operating environments, real-time processing requirements, energy constraints, and the gap between simulated and real-world environments.关键词：unmanned aerial vehicle (UAV);vision-based localization;vision-based navigation;machine vision;image matching168|44|0更新时间：2025-12-08
- 摘要：Unmanned aerial vehicle (UAV)-based multispectral object detection utilizing both visible (RGB) and thermal infrared (T) images, makes all-weather and all-day target monitoring possible, serving critical roles in military and civilian applications. However, due to the complexity of data acquisition and processing, there is currently a lack of publicly available UAV-based RGB-T multispectral object detection datasets, which to some extent limits its research and application. Meanwhile, UAV operational scenarios are characterized by complex and variable conditions, including rapid changes in flight altitude, speed, focal length, and background. So, the captured targets exhibit diverse scales, uneven (dense/sparse) distributions, and category imbalances in images, which presents significant challenges for accurate detection. Furthermore, real-time requirement should be guaranted in applications such as reconnaissance and traffic monitoring. Therefore, it is the key to keep a trade-off between accuracy and speed in the algorithmic design of UAV RGB-T object detector. To address these issues, this paper introduces a large-scale UAV-based RGB-T multispectral dataset named UAV-RGBT, which spans across seasons and day-night cycles, and includes multiple categories and scales. Specifically, UAV-RGBT comprises 20 categories with 5 117 pairs of RGB-T images and over 110 000 annotations, which is conducive to advancing research in UAV-based multispectral object detection algorithms. Moreover, based on the YOLOv8n model, the UAV-based dual-branch multispectral object detection (UAV-DMDet) model is proposed to promote deep fusion of multispectral features through a multi-modal cross-attention fusion module and a multi-modal feature decomposition combination module. This approach achieves a batter trade-off among model parameter size, detection speed, and accuracy. Experimental results demonstrate that the UAV-DMDet model improves the mAP@0.5 on the UAV-RGBT dataset by 3.61% and 11.03% in the visible and thermal modalities, respectively, and enhances the mAP@0.5:0.95 by 0.84% and 6.76%, respectively. On the DroneVehicle dataset, the UAV-DMDet model outperforms the mainstream algorithm I2MDet, with mAP@0.5 and mAP@0.5:0.95 improvements of 2.66% and 12.36%, respectively. Furthermore, with 640
- 摘要：Coverage path planning is one of the key technologies for unmanned aerial vehicle(UAV) swarms in performing the exploration missions such as search and rescue. However, the current research often focuses on the design and optimization of flight paths in a single region, without taking into account quantitatively the effect of UAV capability on region division and start and end point selection in multi-region environment. Meanwhile, most of the existing methods use homogeneous UAV swarms to perform the coverage path planning task, ignoring the ability differences among the UAVs, resulting in a low utilization ratio of swarm resources and much difficulty in adapting to the uncertain changes of tasks and environments. This paper focuses on the coverage path planning problem of heterogeneous UAVs on multiple regions. First, by modeling the heterogeneous UAVs and analyzing the road and energy constraints of the path planning problem, we propose an exact formulation based on mixed integer linear programming to completely search the solution space and to find the best flight roads for UAVs. Then we present an efficient path planning algorithm based on temporal-spatial density clustering to improve the solving efficiency of the coverage path planning problem. The proposed algorithm groups regions according to their densities in time and space, allocates a reasonable group to each UAV, and optimizes the visiting orders of regions and the scan paths in regions, ensuring that the coverage task would be finished effectively. Experimental results show that the proposed method will provide reasonable flight paths for UAVs, and the total flight length and the task completion time can be reduced by 10.55% and 5.47%, respectively.关键词：coverage path planning;heterogeneous unmanned aerial vehicle swarms;temporal-spatial density;density-based clustering;mixed integer linear programming44|19|0更新时间：2025-12-08
- 摘要：Semantic segmentation technology enables fine-grained understanding of complex and diverse scenes and is one of the key technologies to promote efficient and intelligent work of unmanned systems. Large-scale unsupervised semantic segmentation aims to learn semantic segmentation capabilities from a large number of unlabeled images. However, the existing approaches suffer heavily from their noisy self-learned pseudo-labels with poor category and shape representations, leading to low final segmentation accuracy. In this paper, we propose a Pseudo-label Denoising and SAM Optimization (PDSO) approach for large-scale unsupervised semantic segmentation to alleviate the problem mentioned above. Specifically, we first propose a denoising-based feature fine-tuning module, which fine-tunes the pre-trained backbone network with clean image-level pseudo-label samples selected from a large dataset based on a small loss criterion, enabling the network to obtain more robust category representations. To further reduce category noise in pseudo-labels, we propose a clustering-based sample denoising module to discard noisy samples that interfere with clustering based on the category proportion and the distances between samples and cluster centers, thereby enhancing clustering performance. Moreover, we propose a SAM prompt optimization module, which identifies active categories in the image based on clustering distance to filter out noisy targets and uses points and boxes as SAM’s target prompt information to generate expected target masks and refine the edges of targets in pseudo-labels. Our proposed PDSO reaches the mIoU of 45.0%, 26.6%, and 14.5% on the test set of ImageNet-S50, ImageNet-S300, and ImageNet-S919 datasets, respectively, which significantly improves the category accuracy and edge accuracy of the segmented targets.关键词：large-scale unsupervised semantic segmentation;image-level denoising;segment anything model;pseudo-label;clustering21|11|0更新时间：2025-12-08
- 摘要：Multi-drone multi-object tracking aims to predict the tracklets and identities of all targets from videos simultaneously captured by multiple drones, which alleviates the tracking performance degradation when individual drone videos suffer from challenges such as occlusion and cluttered backgrounds. However, the differences in viewpoints and scales of images captured by different drones are usually large, resulting in significant difficulties for aligning and fusing cross-drone features. To address this problem, we propose a novels tracker based on cross-view feature fusion guided by temporal information. It first designs an object-aware alignment network (OAN) that utilizes the tracklet prior during tracking to estimate the transformation relationships between cross-drone frames at previous moments. Then, a temporal-aware alignment network (TAN) is constructed to explore the information of single-drone images in the before-and-after moments to fine-tune the transformation relationship across the images. Finally, based on the cross-drone image transformation relationship estimated by OAN and TAN, this paper presents a cross-drone feature fusion network (CFFN) to fuse the visual information captured by multiple drones, which mitigates the tracking performance degradation in complex scenes. Experimental results on the MDMT dataset show that the proposed TCFNet is more competitive than existing mainstream trackers, exceeding current state-of-the-art model by 2.23, 1.67, and 2.15 percentage points in terms of tracking accuracy, identification F1 score, and multi-device association score.关键词：multi-drone multi-object tracking;temporal information;tracklet prior;cross-view feature fusion;accurate tracking51|20|0更新时间：2025-12-08
- 摘要：In this paper, the scenario of multiple antenna-unmanned aerial vehicle (UAV)-assisted sensing and communication is addressed. Target perception and user communication performance is comprehensively considered, and system cost function is defined as the weighted sum of system energy consumption and user data rate. The optimization problem of communication and sensing scheduling strategy, perception precoding design, and UAV flight trajectory is formulated as a constrained system cost function optimization problem. Due to the highly coupled and non-convex nature of the optimization problem, it is challenging to solve directly. To tackle this problem, the formulated optimization problem is decomposed into three subproblems, i.e., UAV flight trajectory optimization subproblem, communication and sensing scheduling subproblem, and radar perception precoding subproblem, and an iterative nested method is proposed to solve these subproblems. For UAV flight trajectory optimization subproblem, a markov decision process (MDP) is modeled, and a UAV trajectory optimization algorithm is designed based on double deep Q-networks. Given the state of the MDP model, the communication and sensing scheduling subproblem is solved using Lagrange dual transformation and quadratic transformation methods, and the radar perception precoding subproblem is addressed through applying equivalent transformation approaches, i.e., introducing auxiliary variables and converting optimization constraints. Based on the obtained communication and sensing scheduling strategy and radar perception precoding, the reward function of the MDP model is updated and the UAV flight trajectory is determined, so as to achieve the joint optimization of communication and sensing scheduling, perception precoding design, and UAV flight trajectory. The effectiveness of the proposed algorithm is verified through simulations.关键词：unmanned aerial vehicle;integrated communication and sensing;data scheduling;trajectory optimization;precoding49|66|0更新时间：2025-12-08
Intelligent Vision Algorithms for Unmanned Systems
- 摘要：With the development of holographic technology,holographic communication shows a wide range of application prospects in many fields such as education,entertainment and medical care. However,the existing network resource scheduling method is difficult to meet the transmission requirements of holographic communication with large bandwidth,low latency and data synchronization. Smart computing integration network has native in-network computing capability. This allows for computation and optimization of data during packet transmission,thereby reducing transmission latency and network bandwidth pressure. Extends the work from the initial Smart Integration Identifier Network technology,exploring a smart computing integration network for holographic communication. It introduces an identifier mapping method based on service identifiers,resource chain identifiers,network function identifiers,and network component identifiers,enabling adaptive scheduling of holographic services to diverse resources such as computing,storage,and forwarding. Based on the identifier mapping method,this paper further designs a holographic communication smart computing integration network system. Experimental results show that,compared to traditional edge computing methods,the smart computing integration network can reduce the network bandwidth pressure by 60% and transmission latency by 45% for holographic communication services,while also offering more flexible network resource scheduling capabilities.关键词：smart computing integration network;in-network computing;holographic communication;network resource scheduling24|42|0更新时间：2025-12-08
- 摘要：As one of the typical equipment of cyber-physical systems (CPS), UAVs are easy to use, have low requirements for the working environment and strong flexibility, and have been widely used in agriculture, industry, military and other fields. Among them, the flight control system is the core basic service of UAV, which ensures the effective implementation of UAV telemetry perception, communication coverage, surveying, mapping and disaster relief applications. However, the changeable physical environment and complex functional structure make it easy to introduce various software security problems in the development process of the UAV flight control system, resulting in serious problems such as hijacking, crashing, and loss of control of the UAV. How to detect the security of the UAV flight control software system has become very important. Most of the existing UAV anomaly detection technologies rely on the input of digital world construction, and it is difficult to find the problem of UAV logic security in time, so this paper proposes a security detection method for UAV flight control software that supports physical interaction, combines static and dynamic analysis methods, and combines fuzzing testing methods to test the security of UAV flight control software, the results show that the method can detect the safety of UAV flight control tasks with a high coverage rate of 97%, and extract UAV feature data according to the test resultsBased on the feature data, the machine learning method is used to train a double anomaly detection model, and by comparing with the existing detection methods on multiple datasets, the proposed method finds the abnormal condition of the UAV with an accuracy rate of 97.5%, and effectively detects the known safety problems in the UAV flight control software system.关键词：unmanned aerial vehicle flight control;fuzzing;software security;anomaly detection;security testing18|23|0更新时间：2025-12-08
- 摘要：A substrate integrated waveguide (SIW) periodic leaky-wave antenna (PLWA) with increased gain and continuous beam scanning through broadside is proposed by loading slow-wave (SW) structures. Slow wave structure in the form of periodic blind via-holes with loaded patches decelerates the phase velocity of the electromagnetic wave traveling in SIW, reducing the guided wavelength by 50%. Compared with normal SIW PLWA, the distance between the adjacent radiating slots in the proposed SW-SIW PLWA is decreased by half, allowing twice as many radiating slots in SW-SIW PLWA with the same length. Therefore, the radiation efficiency and the gain of SW-SIW PLWA can be significantly increased. Furthermore, the loaded patches of slow wave structure under the radiating slots are extended to improve the impedance match of the radiating slots and suppress the reflected wave, so the open-stopband (OSB), which is a common drawback of the PLWA, is suppressed. In consequence, the radiation beam can scan from backward to forward direction continuously. A prototype of the proposed SW-SIW PLWA is manufactured and measured, the scanning angle of the proposed PLWA reaches 72.7° during the operating frequency range of 13.4~15.4 GHz with maximum gain of 8.47 dBi. The measured results agree well with the simulations.关键词：periodic leaky-wave antenna;substrate integrated waveguide;slow wave;open-stopband;continuously scanning22|16|0更新时间：2025-12-08
- 摘要：Breathing, as a crucial physiological process for sustaining life, is closely related to various respiratory diseases such as sleep apnea and asthma. To meet the increasing demand for health monitoring, this paper innovatively proposes a dual-channel wearable MEMS (Micro-Electro-Mechanical Systems) respiratory monitoring microsystem integrated with a flexible nasal expander. This microsystem incorporates a flexible nasal expander, a respiratory sensor, and a signal processing module, enabling continuous real-time monitoring of airflow within the nasal cavity. The sensor’s sensitive element adopts a folded metal resistor structure, deposited on a glass substrate through planar MEMS technology, utilizing the thermoresistive effect to achieve signal measurement. When embedded in the flexible nasal expander, the sensor can simultaneously monitor breathing signals from both sides of the nasal cavity, making it especially suitable for long-term continuous monitoring. Signal simulation and performance testing results demonstrate that the sensor exhibits excellent sensitivity, response speed, and anti-interference capability. In tests simulating respiratory conditions such as sleep apnea and asthma, the sensor accurately differentiates between normal and abnormal breathing patterns, supporting further analysis of various respiratory diseases. Based on this, the paper develops a dual-channel wearable MEMS respiratory monitoring device integrated with a flexible nasal expander, aimed at continuous, real-time, and long-term respiratory monitoring, particularly suitable for abnormal breathing screening and health monitoring during sleep. Additionally, this system captures changes in the nasal cycle, providing new data dimensions for in-depth analysis of breathing patterns and physiological rhythms, highlighting its potential application value in long-term health management.关键词：respiratory monitoring;micro-electro-mechanical systems;wearable devices;thermoresistive effect;respiratory diseases;respiratory sensor13|11|0更新时间：2025-12-08
- 摘要：Through first-principles calculations, we systematically investigate the adsorption performance of molecules such as NO₂, NO, CO, CO₂, CH₄ and SO₂ on Ge-doped phosphorene by optimizing geometric structures and analyzing corresponding electronic properties. The adsorption positions of these molecules on the material surface are identified, and the most stable adsorption configurations are determined. The study reveals that Ge-doped phosphorene exhibits strong adsorption capabilities toward molecules such as NO₂, NO, CO, CO₂, CH₄ and SO₂, indicating its potential application prospects in gas molecule adsorption. Furthermore, the influence of Ge doping on the electronic structure of the material is explored. Results demonstrate that Ge doping introduces new energy levels, which modify the conductive properties of the material and consequently regulate the interactions between adsorbed molecules and the substrate. This provides a theoretical foundation for understanding the adsorption mechanisms of gas molecules on Ge-doped phosphorene. In summary, this work highlights the superior performance of Ge-doped phosphorene in adsorbing molecules such as NO₂, NO, CO, CO₂, CH₄ and SO₂, while offering theoretical insights into its potential applications in gas adsorption and related fields. These findings hold significant implications for developing novel high-efficiency adsorption materials to address environmental and energy challenges.关键词：phosphorene;First-principles;gas sensors;doping;gas sensing property;two dimensional materials17|10|0更新时间：2025-12-08
- 摘要：It is crucial to monitor ethanol in real time due to the safety risks posed by its high volatility and flammability. However, current methods for improving the performance of SnO2 ethanol sensors often hinder the miniaturization of devices. To address this, the paper designs an intrinsic SnO2 ethanol sensor with a field-effect transistor structure and employs magnetron sputtering to fabricate the sensitive film. The study systematically investigates the influence of gate voltage on the gas-sensing performance of the sensor. Experimental results indicate that the SnO2 sensor prepared by sputtering is an n-channel depletion-mode device. Gas-sensing tests reveal significant differences in the sensor’s response under different operating gate voltages: at a gate voltage of 10 V, the current change of the sensor in 100 ppm ethanol is 2.40 times; while at a gate voltage of -30 V, the channel current change is significantly enhanced to 3.42 times, representing a 42% improvement compared to 10 V. Further investigation shows that the gas-sensing properties of SnO2 arise from the modulation of carrier concentration in the channel by the surface adsorption of ethanol molecules. This effect is significantly enhanced under negative gate voltage but suppressed under positive gate voltage. However, a positive gate voltage of 10 V induces more electrons in the channel, effectively accelerating the adsorption and desorption processes of ethanol. As a result, the sensor’s response and recovery times to 100 ppm ethanol are reduced to 8 s and 17 s, respectively, demonstrating faster dynamic characteristics. The study’s findings indicate that the degree and rate of ethanol vapor reaction on the SnO2 surface are significantly regulated by the sensor’s gate voltage. This research provides a new approach for optimizing the gas-sensing performance of SnO2 sensors and contributes to advancing their application in miniaturized, fast-response, and high-precision gas-sensing detection.关键词：pristine SnO2;filed effect transistor;gate voltage tunable;ethanol sensor29|15|0更新时间：2025-12-08
- 摘要：With the rapid development of industrialization and urbanization, the importance of environmental monitoring is becoming more and more prominent. However, traditional monitoring methods are limited by high costs, difficult layout and maintenance challenges, making it difficult to achieve comprehensive and real-time monitoring. Crowd Sensing, an emerging environmental monitoring method, utilizes widely used highly intelligent devices and integrated sensors for large-scale collection and real-time transmission of environmental data. However, existing studies seldom consider data privacy protection, work balance, and system cost at the same time, which makes it difficult to achieve the expected results in practical applications. To solve this practical problem, this paper proposes a low-cost and high-efficiency method that can be applied to crowd sensing for environmental monitoring (Adaptive Federated Learning based Crowd Sensing algorithm for Environmental Monitoring, AFL-CSEM). Specifically, we first consider the challenges of resource constraints, device heterogeneity, and non-independent and homogeneous distribution of data in the system, and model the system by combining crowd sensing and federated learning techniques, and train the model locally on user’s devices, sharing only the model parameters to effectively protect data privacy. Then, the convergence analysis of the system is carried out, and the convergence bounds of the crowd sending algorithm based on federated learning are obtained for non-independently and identically distributed data distributions. Then, in order to reduce the impact of device heterogeneity, based on the results of the convergence analysis, an adaptive control method is designed to dynamically adjust the local update frequency and batch size to adapt to the heterogeneous and dynamic monitoring environment. By comparing on real datasets, all the experimental results consistently prove the effectiveness of the proposed algorithm in this paper, and the AFL-CSEM algorithm improves the efficiency and accuracy of model training while reducing the computation and communication overhead and lowering the economic cost. It provides a novel and informative solution for environmental monitoring in resource-constrained edge computing environments.关键词：environmental monitoring;crowdsensing;federated learning;adaptive algorithm;convergence analysis22|9|0更新时间：2025-12-08
- 摘要：This paper proposed a protocol of semiquantum private comparison (SQPC) based on entanglement swapping of GHZ-like state and Bell state, which allows the classical participants to compare the equality of their secret message under the help of a semi-honest third party (TP). TP is allowed to misbehave but cannot collude with anyone else. This paper provides a detailed proof of the protocol’s complete robustness against external eavesdroppers’ attacks, and analyzes its security against dishonest internal participants. This paper also conducted experimental simulations on the flow and output correctness of the protocol using IBM’s Qiskit. In addition, the security of the proposed protocol is confirmed and it can effectively prevent various kinds of attacks.关键词：semiquantum private comparison;entanglement swapping;GHZ-like state;Bell state;Qiskit13|9|0更新时间：2025-12-08
- 摘要：Existing malware detection methods suffer from inadequate extraction of sample features, excessive reliance on domain expert knowledge, and operational behavior monitoring, significantly impacting detection and classification performance. To address these issues, we propose a malware detection method based on multidimensional dynamic weighted alpha image fusion and feature enhancement. Standardized sample sets are obtained through invalid sample cleaning and outlier processing. High-quality fused image samples are then generated using a three-channel image generation and multidimensional dynamic weighted alpha image fusion method. The puppet optimization algorithm is employed for data reconstruction to mitigate the impact of data class imbalance on detection results, and image enhancement is performed on the reconstructed data samples. A spatial attention enhancement network based on dual-branch feature extraction and fusion channel information representation is used to extract and enhance image and text features, thereby improving feature representation capabilities. The enhanced image and text features are fused using a weighted fusion method to achieve malware family detection and classification. Experimental results show that the proposed method achieves a malware detection classification accuracy of 99.72% on the BIG2015 dataset, representing an improvement of 0.22~2.50 percentage points over existing detection methods.关键词：malware detection;image fusion;puppet optimization algorithm;dual-branch feature extraction;data reconstruction;Feature enhancement38|14|0更新时间：2025-12-08
- 摘要：Mutation testing is a pivotal technology for enhancing software quality by injecting mutation operators to generate mutant programs that mimic potential defects in software. The substantial number of mutants and the associated execution costs, however, limit the advancement and industrial application of mutation testing. The selection of effective mutation operators is a primary strategy for reducing the volume of mutants. Addressing distributed memory parallel programs, this paper introduces an evaluation criterion for the effectiveness of mutation operators. Mutants are categorized into three types: stubborn mutants, crash mutants, and equivalent mutants. Drawing on the influence of various mutants on the quality of test data, we establish a criterion for evaluating the effectiveness of mutation operators and analyze the performance of different mutation operators. Experimental results demonstrate that the proposed evaluation criterion enables the selection of suitable mutation operators, leading to the generation of a higher number of valid mutants and a minimized number of invalid mutants. Consequently, while preserving the effectiveness of mutation testing, the average number of mutants is decreased by 22.61%, thereby enhancing the efficiency of mutation testing.关键词：distributed memory;parallel programs;mutation testing;mutation operators;mutants24|10|0更新时间：2025-12-08
- 摘要：In spatial crowdsourcing, task allocation is a crucial prerequisite for subsequent location-aware data collection. To tackle potential location privacy breaches, researchers often adopt geo-indistinguishability. Existing approaches that satisfy Geo-I are often designed for one-to-one scenarios, while implicitly assume that workers can perform any task, and they often focus on minimizing the average travel distance, rather than maximizing the number of task allocation. Furthermore, these studies often incorporate the planar laplacian mechanism to achieve Geo-I. However, due to the randomness and unbounded nature of PL, it can result in excessive noise in the location data uploaded by workers, significantly deteriorating the utility of task allocation. This can lead to either long distances or unassigned tasks. To address these problems, we propose MONITOR (Many-to-many task allOcation under geo-iNdIsTinguishability for spatial crOwdsouRcing), a new privacy-preserving task allocation approach for many-to-many scenario. The general idea of MONITOR is to upload the distances from each worker’s true location to the obfuscated preferred tasks’ locations instead of uploading each obfuscated worker’s location. In MONITOR, to collect the distances for subsequent task allocation, we design an obfuscated distance collection method, called GroCol (Group-based obfuscated distance Collection). To improve the utility for task allocation, we develop a parameter independent obfuscated distance comparison method called ParCom (Parameter-free obfuscated distance Comparison). To illustrate the effectiveness of MONITOR, we first theoretically analyze its privacy guarantee, task utility, and computational complexity. We then empirically show on two real-world datasets and one synthetic dataset that MONITOR share similar results to that of non-private task allocation about the number of assigned tasks, and reduce the average travel distance by more than 20% compared to the baseline approaches.关键词：spatial crowdsourcing;task allocation;privacy protection;geo-indistinguishability;average travel distance14|18|0更新时间：2025-12-08
- 摘要：Anomaly detection has been widely studied and applied to various visual scenes. Recently, the mainstream unsupervised anomaly detection schemes are usually based on distillation methods and reconstruction methods. However, they still have some limitations. In distillation model, the student network can usually learn the strong representation ability of the teacher network, thus can not represent differently for the abnormal regions. In reconstruction model, the encoder-decoder model can easily learn a restoration shortcut and recover features indiscriminately. To address the above challenges, we propose
- 摘要：RGB and Thermal infrared (RGBT) tracking is a multi-modal object tracking method that integrates different information from visible light and thermal infrared sensors. This method aims to overcome the limitations of single sensor in a specific condition and increase the robustness and accuracy of object tracking by fusing data from multiple sensors. However, the majority of RGBT tracking methods in use today directly fuse features extracted from thermal infrared and visible light images, ignoring the homogeneity and heterogeneity of the two modalities. In addition, RGBT tracking is often affected by multiple challenging factors such as objects fast motion, scale variation, illumination variation, thermal crossover, and occlusion. Existing work often focuses on a single model to solve all challenges simultaneously, which requires highly complex model and extensive training data. This paper proposes a novel network called CMHHNet (facing different Challenges and combining Multi-modal Homogeneous and Heterogeneous information separation and integration Network) for RGBT tracking. In this network, a challenge-aware module is deployed in each layer of the backbone to fuse the visible light and thermal infrared features from two different modalities under each challenge separately, and adaptively aggregate the fused features under all challenges. In addition, an attention enhancement module and a multi-scale auxiliary module are added to strengthen the features that the backbone network has extracted. Finally, according to the homogeneity and heterogeneity of thermal infrared and visible light, their unique and common features are extracted separately and adaptively fused. Extensive experiments on GTOT, RGBT234 and LasHeR datasets demonstrate that the tracker proposed in this paper shows quite strong competitiveness compared with existing RGBT tracking methods.关键词：RGBT tracking;challenge-aware;separation of homogeneous and heterogeneous information;adaptive aggregation;attention mechanism;multiscale features13|8|0更新时间：2025-12-08
- 摘要：Images captured in low-light scenes are susceptible to multiple degradations such as darkness, noise, and blur, resulting in poor visibility and visual perception. Multi-degraded low-light image enhancement poses challenges to existing image enhancement methods as follows: on the one hand, low-light image enhancement or deblurring methods cannot handle all three types of degradation, and the effect of the combination strategy is limited by the increased computational cost and error accumulation. On the other hand, the existing multi-degraded low-light image enhancement method adopts the strategy of enhancing brightness first and then removing blur, and this sequential processing manner increases the risk of losing feature cues and is not conducive to detail recovery. To cope with the above challenges, this paper proposes the progressive edge-aware interactive enhancement network (PEIE-Net), which reduces the loss of feature details by designing a step-by-step enhancement process. Specifically, our network consists of an image enhancement branch and an edge information prediction branch. In each enhancement stage of the image enhancement branch, a self-attention modulation prediction module is designed to extract the global information, which is used for adaptive modulation in the channel modulation module and multi-scale restoration module. In the edge information prediction branch, the spatial-frequency domain feature transformation module is developed to extract the edge perceptual information. The edge perceptual information is used to predict the edges of high-quality images while also fused with the image enhancement features, simulating the interaction between different perceptions within the human visual system. In addition, we propose scene brightness estimation loss to coordinate the multiple progressive enhancement stages. Experiments on synthetic and real datasets demonstrate the effectiveness and sophistication of our method for enhancing low-light, noisy, and blur-degraded images, and can be used for low-light image enhancement and super-resolution tasks.关键词：image enhancement;multi-degraded image;low-light enhancement;deblurring;feature modulation25|9|0更新时间：2025-12-08
- 摘要：To address the issues of low spatial resolution and susceptibility to noise in traditional single-modality brain-computer interface (BCI) technologies based on electroencephalography (EEG), an increasing number of studies have focused on BCI research that combines EEG signals with functional near-infrared spectroscopy (fNIRS) signals. However, integrating these two heterogeneous signals poses challenges. This paper proposes an innovative end-to-end signal fusion method based on deep learning and evidence theory for motor imagery (MI) classification. The spatiotemporal feature information of EEG signals is extracted using dual-scale temporal convolution and depth wise separable convolution, with a hybrid attention module introduced to enhance the network’s ability to perceive important features. For fNIRS signals, spatial convolution across all channels explores activation differences between different brain regions, while parallel temporal convolution and gated recurrent unit (GRU) capture richer temporal feature information. During the decision fusion stage, the decision outputs obtained from decoding each signal are first utilized to estimate uncertainty using Dirichlet distribution parameter estimation. Subsequently, Dempster-Shafer theory (DST) is employed for dual-layer reasoning, effectively merging evidence from the two basic belief assignment (BBA) methods and different modalities to obtain the decoding results. The proposed model is evaluated on the publicly available TU-Berlin-A dataset, achieving an average accuracy of 83.26%, which represents a 3.78 percentage points improvement compared to the state-of-the-art research. This provides new ideas and approaches for fusion studies based on EEG and fNIRS signals.关键词：hybrid brain-computer interface (BCI);motor imagery (MI);deep learning;Dempster-Shafer theory (DST);functional near-infrared spectroscopy (fNIRS) signal;electroencephalography (EEG) signal48|14|0更新时间：2025-12-08
- 摘要：Recognition of 3D multimodal positron emission tomography/computed tomography (PET/CT) lung tumor using deep learning is an important research area. In medical images of lung tumors, the spatial shape of lesions is irregular and the boundary between the lesions and the surrounding tissues is blurred, which makes it difficult for the model to fully extract tumor features, and the computational complexity of the model is higher in three-dimensional tasks. To solve the above problems, a cross-modal Light-3Dformer 3D lung tumor recognition model is proposed in this paper. The main contributions of this paper are as follows. Firstly, the backbone network extracts PET/CT image features, and the auxiliary network extracts PET image features and CT image features. Multi-modal feature enhancement and interactive learning are realized by lightweight cross-modal collaborative attention. Secondly, Light-3Dformer module are designed. In this module, Updating the 2 times matrix multiplication operation of Transformer to the linear element multiplication operation of Lightformer; The cascade Lightformer structure is designed, the output feature map of the cascade Lightformer structure and the initial input feature map are fused, through parallel and deep and shallow feature fusion, lightweight and rich gradient information can be realized; Designing with parameter less attention, this structure can enhance the ability of lung tumor feature extraction from three aspects: channel, space, and tomography image. Thirdly, lightweight cross-modal collaborative attention module (LCCAM) is designed, which can fully learn the cross-modal advantage information of 3D multi-modal images and carry out interactive learning of deep and shallow features. Finally, ablation experiments and comparative experiments. In the self-built 3D multi-modal data set of lung tumor, the accuracy and area under the curve (AUC) values of the model are 90.19% and 89.81%, respectively, under the premise of optimal computation and running time. Comparing with the 3D-SwinTransformer-S model, the computation quantity is reduced by 117 times, and the calculation quantity is reduced by 400 times. The experimental results show that the model can better extract multi-modal information of lung tumor lesions, which provides a new idea for lightweight and multi-modal interaction of deep learning 3D models.关键词：lung tumor;multimodal images;Transformer;Light-3Dformer;light cross-modal collaborative attention9|11|0更新时间：2025-12-08
- 摘要：An image classification algorithm based on coordinate importance pooling and decoupled class alignment distillation is proposed to improve the image classification accuracy of convolutional neural networks while achieving network lightweighting. Firstly, a coordinate importance pooling module is designed and embedded it into ResNet34, in order to fully utilize the positional information of image pixels to enhance the ability to discriminate important features. Secondly, BlurPool is used to mitigate the impact on network performance due to shift equivariance during down-sampling, and to construct the teacher network. Finally, the decoupled class alignment distillation algorithm was constructed to efficiently migrate image classification knowledge from the teacher network to the lightweight MobileNetV3 network, which considers the knowledge of target and non-target class separately and introduces correlation information between the class. The experimental results on different datasets showed that the proposed teacher network effectively improves the classification performance, and the distillation-trained student network achieves superior overall performance than other networks of the same magnitude, making it better applicable to practical scenarios with limited computational and storage power.关键词：image classification;lightweight;knowledge distillation;ResNet34;coordinate importance pooling13|11|0更新时间：2025-12-08
- 摘要：In the end-to-end text recognition of complex natural scenes, because text and background are difficult to distinguish, the location information detected by text and the semantic information recognized do not match, and the correlation between detection and recognition cannot be effectively utilized. In response to this problem, this paper proposes a multi-party synergetic information with dual-domain awareness text spotting (MSIDA). By enhancing text region features and edge textures, the synergies between text detection and recognition features are utilized to improve end-to-end text recognition performance. Firstly, a dual-domain awareness (DDA) module integrating text space and direction information is designed to enhance the visual feature information of text instances. Secondly, a multi-party explicit information synergy(MEIS) is proposed to extract explicit information from coding features and generate candidate text instances by matching and allocating the position, classification and character multi-party information used for detection and recognition. Finally, cooperative features guide learnable query sequences through decoders to obtain text detection and recognition results. Compared to the latest decoder with explicit points solo (DeepSolo) method, on the Total-Text, ICDAR 2015 and CTW1500 datasets, the accuracy of MSIDA improved respectively by 0.8%, 0.8% and 0.4%. The code and datasets are available at https://github.com/msida2024/MSIDA.git.关键词：computer vision;scene text images;text detection;text spotting;feature information synergy14|12|0更新时间：2025-12-08
- 摘要：The privacy and security of speech are fundamental to both national and personal information security. To protect users’ speeches from being eavesdropped on, ultrasonic microphone jammers are widely utilized. These jammers utilize the nonlinear characteristics of ultrasound in digital recording devices to inject noise into microphones efficiently and cost-effectively, without disrupting normal communication or human hearing. However, existing microphone jammers are vulnerable. They merely introduce simple noise to mask speeches. As a result, eavesdroppers can employ advanced denoising techniques to recover speech information, posing a significant threat to speech privacy and security. Moreover, existing jammers have primarily been designed for English speech, limiting their applicability to Chinese speech. Therefore, there is an urgent need for privacy protection for Chinese speech. To enhance the security and adaptability of ultrasonic microphone jammers, this paper introduces a robust jammer for Chinese speech privacy protection. Based on the unique characteristics of Chinese phonetics, we design a coherent noise generation algorithm, which produces real-time ultrasound noise intimately coupled with the protected speech signal. This noise is designed to be difficult for adversaries to separate from the speech, ensuring that any attempts at eavesdropping will be frustrated. Comprehensively considering the capabilities of the potential adversaries adversary, our proposed jammer realizes the robust protection against eavesdropping. The generated noise cannot be removed by adversaries using state-of-the-art denoising techniques and is imperceptible to human hearing. Thereby, we comprehensively safeguard speech privacy and security. We develop a prototype of the proposed ultrasonic microphone jammer to validate its effectiveness. Experimental results demonstrate that over 90% of protected speeches remain unrecognizable to adversaries within a range of 6 meters under the protection of the proposed jammer, even if the adversary adopts state-of-the-art denoising techniques. Therefore, we provide robust technical support to protect Chinese speech privacy.关键词：speech security and privacy;Internet of Things security;mobile security;Chinese phonetics;ultrasonic non-linearity18|12|0更新时间：2025-12-08
- 摘要：Alzheimer’s disease (AD) is a chronic neurodegenerative disease, and its accurate classification is advantage to achieve early diagnosis of AD so as to take timely treatment and intervention. In this paper, a novel method on AD Classification utilizing graph neural network with nearest neighborhood aggregation (GraphNAGE) is proposed. Firstly, the graph data modeling is performed to represent AD samples as graph data. By feature selection method based on mutual information (MI), the high-importance volume features are selected from the 114 dimensional volume features of cerebral cortex and subcortical regions of interest (CCS-ROI) in the sample, and used for node modeling. Meanwhile, a relationship modeling method based on similarity measurement, modeling the relationships between samples using high importance volume features, genetic genes, demographic information, and cognitive scores, is presented. Subsequently, the graph neural network with nearest neighborhood aggregation is constructed. For each node in the graph data, the nearest neighbor sampling is performed based on the weights of edge related to it. Then, the sampled data of neighboring nodes and central node are aggregated using the mean aggregation method. At last, a full-connected layer and a softmax layer are used to implement AD classification. The proposed AD classification method is evaluated on the Alzheimer’s disease prediction of longitudinal evolution (TADPOLE) dataset. The accuracy (ACC),
PAPERS
- 摘要：The measurement and control of advanced air vehicle requires the realization of multiple functions such as telemetry, remote control, communication, and tracking. Traditionally, it is generally composed of multiple wireless transceiver systems and discrete antennas. The contradiction between its volume, weight, cost, installation, etc. and the limited resources of the air vehicle is becoming increasingly prominent. The antenna aperture synthesis enables a single multi-functional antenna aperture to perform the functions of multiple dedicated antenna apertures. This greatly reduces the number of antenna apertures. It also significantly eases the pressure on the antenna aperture layout on the air vehicle platform, offering a new way to enhance the system-level electromagnetic compatibility. This paper systematically elaborates on the technical route of antenna aperture synthesis for air vehicle measurement and control communication. It focuses on introducing the multi-band and multi-polarization antenna technology for the synthesis of multiple discrete antennas, the diplexer antenna technology for the synthesis of transmitted and received antennas, the shared-aperture antenna technology for the synthesis of multiple antennas in the same aperture, and the coupling suppression technology for the integration of the same-frequency antenna array. At the same time, combined with the working characteristics of the software-defined radio system, it analyzes the advantages and feasibility of the application of the software-defined radio system in the air vehicle measurement and control communication system. Finally, this paper looks forward to the development of the antenna aperture synthesis technology for air vehicle measurement and control communication and puts forward the possible development directions of the antenna aperture synthesis technology in the development of the air vehicle measurement and control communication system.关键词：telemetry, tracing and control;antenna aperture synthesis;software defined radio52|19|0更新时间：2025-12-08
- 摘要：With the continuous construction of new information infrastructures, multi-dimensional array signal processing plays a fundamental role in the filed of radar, wireless communication, remote sensing and so on. Multidimensional array signals contain rich spatial/temporal/frequentiol/polarization parametric information, offering great economic and social values. To deal with the problem of structural information loss inherent in traditional vector/matrix models, the tensor algebra has been adopted to effectively retrieve multi-dimensional signal features. However, as the dimension of signals increases, the tensor signal volume following the Nyquist sampling theorem exponentially expands. Unfortunately, computation resources of the system are approaching the physical limit, resulting in computational overload and high latency. Concerning these issues, the sparse sensing theory has been developed to exploit the spatial sparsity of signals for sub-Nyquist processing. The extension from one-dimensional sparse sensing to multi-dimensional sparse sensing becomes a promising solution to efficient tensor signal processing. Meanwhile, by imposing structured sparse sensing paradigm such as coprime and nested sensing, the performance of the system can be enhanced via augmented coarray signal processing. Thus, to pursue the high economy of multi-dimensional array signal processing, this paper endeavors to the research on Structured Sparse Tensor Signal Processing for Sensor Arrays. In particular, the paper introduces the statistical theory of sub-Nyquist tensor signals. By deriving the augmented coarray tensor model and devising the corresponding strategy of source identifiability enhancement, this theory facilitates Nyquist matching in the virtual domain and underdetermined parameter estimation. Based upon this theory, this paper introduces a coarray tensor completion algorithm for sparse array DOA estimation, exploiting the full information of the discontinuous virtual array to achieve high accuracy and resolution. Meanwhile, this paper introduces a coprime tensor weights optimization algorithm for sparse array beamforming, which yields a beampatten with a sharper mainlobe and lower sidelobes, and increases the output signal-to-interference-plus-noise ratio. Furthermore, this paper introduces a resource-efficient tensorized neural network for robust sparse tensor signal processing, which compensates the performance deterioration for the model-driven methods in non-ideal conditions by efficiently learning tensor signal features.关键词：multi-dimensional array signal processing;tensor signal processing;structured sparse sensing;direction-of-arrival estimation;beamforming33|20|0更新时间：2025-12-08
SURVEYS AND REVIEWS
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0