In distributed electromagnetic countermeasure scenarios, our jamming transmissions directed against adversarial targets may inadvertently interfere with our own co-frequency electromagnetic systems. To address this issue, a distributed co-frequency jamming cancellation architecture and method are proposed. The method models the time-delay errors and residual frequency offsets between the distributed jamming nodes and the authorized receiving nodes, and solves for the model coefficients using the recursive least squares algorithm. This approach effectively accounts for high-dimensional channel and time-frequency variations. The reference jamming signal is then used for the reconstruction and cancellation of the distributed jamming. Simulation results demonstrate that under normalized timing errors of 0.2 and normalized residual frequency offsets of 10^-4, the proposed distributed co-frequency jamming cancellation method achieves a 13.1 dB improvement in jamming cancellation capability over the single-node approach, with residual jamming approaching the noise floor.

Natural language description-driven object tracking refers to guiding the visual tracking task through natural language descriptions, and fusing textual descriptions and image visual information to realize the model's perception and understanding of the world “like a human”. With the development of deep learning, new methods in the field of natural language description-driven visual tracking are emerging. However, most of the existing methods are limited to two-dimensional space and fail to fully utilize the position information in three-dimensional space, and thus are unable to naturally perceive the world in three dimensions as humans do. Most of the existing 3D object tracking tasks rely on expensive sensors and have limitations in data acquisition, which makes 3D object tracking even more complicated. To address the above challenges, this paper proposes a new task of Natural Language-driven Object Tracking in 3D(NLOT3D) in monocular view and constructs the corresponding dataset, NLOT3D-SPD. In addition, this paper designs an end-to-end NLOT3D-TR(Natural Language-driven Object Tracking in 3D based on Transformer) model, which fuses visual and textual cross-modal features and achieves excellent experimental results. This paper provides a comprehensive benchmarking of the NLOT3D task with several comparative experiments and ablation studies, providing strong support for further development in the field of 3D object tracking.

Deep learning-based Synthetic Aperture Radar (SAR) target recognition methods are widely used in military reconnaissance and disaster monitoring. However, Deep Neural Networks (DNNs) are vulnerable to adversarial attacks, which compromise the reliability of model decisions. Existing black-box adversarial attack methods for SAR images face challenges such as high-dimensional parameter design and perceptible perturbations. To address these issues, a frequency-domain multi-objective optimization-based adversarial attack method is proposed. By transforming SAR images from the spatial domain to the frequency domain via 2D Discrete Fourier Transform, the method reduces perturbation design complexity and modifies a single frequency component to generate texture-like perturbations in the spatial domain. A hypervolume metric-guided multi-objective evolutionary algorithm is integrated to balance attack performance and visual stealthiness. Experimental results demonstrate that, for the T62 category, the adversarial samples generated by our method achieve misclassification confidence rates of more than 90.39%, 71.43%, 44.28% on VGG16, AConvNet, and YOLO series models, respectively. Additionally, the similarity between adversarial and original images exceeds 99% across all cases, providing effective technical support for security and robustness evaluation of SAR imaging systems.

Utilizing properties of group algebras over finite fields, we construct a class of Hermitian linear complementary dual (LCD) 2-quasi-abelian codes. Employing the structure theorem for group algebras over finite fields, we explicitly determine the number of such codes. By investigating the enumeration of codes within this class that possess small relative minimum weights, we demonstrate that the class of Hermitian LCD 2-quasi-abelian codes over any finite field is asymptotically good.

Unsupervised domain adaptive (UDA) person re-identification (Re-ID) seeks to leverage labeled source domain data to address the task of unsupervised Re-ID in unlabeled target domain data. Recently, contrastive learning has attracted attention in this field. However, current methods suffer from small differences in positive sample pairs and overlook biases in negative proxy sampling. To resolve these challenges, this paper presents a progressive hybrid contrastive learning (PHCL) method. In each training epoch, the PHCL method divides the unlabeled dataset into clustered samples with pseudo-labels and un-clustered independent instances through two steps: clustering and progressive refinement. Based on the clustering results, PHCL implements contrastive learning at two different levels: to learn intra-category similarity through bringing together similar samples within the same cluster (target domain) or identity label (source domain) and explores inter-instance discrimination by applying repulsion among un-clustered individual instances. Moreover, the PHCL method generates positive proxies for anchor samples through nearest neighbor mining, increasing the differences among positive sample pairs to learn richer semantic information. Additionally, the PHCL method performs debiasing in the negative proxy sampling process, mitigating the adverse impact of false negative proxies on model training. Experimental results show that the PHCL method achieves mean average precision (mAP) of 85.9% and 42.3% on the Market-1501 and MSMT17 datasets, respectively, which are improvements of 4.3 percentage points and 13.5 percentage points over the baseline model. These results validate the efficacy of the PHCL method for UDA Re-ID.

The widespread adoption of complex machine learning models across diverse industries has significantly increased the demand for model interpretability. The counterfactual explanation is a crucial post-hoc explanation method. However, traditional approaches often combine multiple objectives into a single objective optimization problem, leading to difficulties in weight assignment and reconciling conflicting objectives. Furthermore, existing methods also suffer from low computational efficiency, degraded prediction accuracy, and insufficient global explanations when dealing with high-dimensional, redundant, and noisy data. To address these issues, this article proposes a comprehensive causal multi-objective counterfactual explanation method with feature selection (CCE-FS). CCE-FS first employs the maximal information coefficient (MIC) to select key features, thereby enhancing prediction accuracy and global explanatory power. It then formulates the counterfactual search as a multi-objective optimization problem, effectively balancing the relationships between multiple objectives. Domain-specific causal relationships are incorporated as constraints to ensure the generated counterfactuals are realistic and plausible. Additionally, CCE-FS provides visual feature effect analysis to enhance user understanding and reveal potential model biases. Experiments conducted on the Statlog dataset demonstrate that CCE-FS significantly improves the validity, normality, and sparsity of counterfactual samples through feature selection, achieving a 46.3% enhancement in proximity for continuous features. Further validation on the Adult-Income and COMPAS datasets confirms that CCE-FS outperforms existing methods in causal consistency, data distribution reasonableness, and proximity of continuous features. These results highlight CCE-FS's superior explanatory capabilities and greater application potential.

Lower limb exoskeletons require the capability to identify the user’s lower-limb motion intentions to provide support during daily activities. However, existing research rarely focuses on predicting locomotion modes that provide user intention for new subjects. To bridge this gap, this study proposes a novel method for lower-limb locomotion mode prediction based on multi-sensor signal fusion and transfer learning. The study first designs a prediction model that utilizes Long-Short Term Memory (LSTM) networks to extract pattern features from Surface Electromyography (sEMG) signals. These sEMG features are then fused with joint angle features to predict lower-limb locomotion modes. Considering the inter-subject variability in physiological signals, the method employs a two-step training process using transfer learning. First, the model is pre-trained on a source domain dataset. Next, the weights of the sEMG feature extractor are frozen, and the fully connected layers are fine-tuned using a target domain dataset. Experimental data were collected from subjects performing both normal walking and exoskeleton-wearing walking. Experimental results with a prediction time of 100 ms demonstrate that the proposed method enhances motion pattern prediction accuracy by 9.53% during free walking and by 8.29% during exoskeleton-wearing walking for new subjects. These results suggest that the proposed approach can improve locomotion mode prediction accuracy for new subjects, thereby ensuring reliable human motion intention prediction in lower-limb exoskeletons.

Voting is an important decision-making method in modern society. In this paper, we propose an efficient multi-centre quantum-secure voting scheme using quantum walk and semi-quantum techniques. This scheme consists of multiple voters, multiple quantum centres, etc. This scheme uses semi-quantum techniques to reduce the equipment cost and facilitate the implementation; Multiple quantum centers are computed in parallel, and the combination of ring and star structures reduces the communication pressure on the central nodes, making voting and vote counting more efficient and suitable for scenarios with a large number of people voting; When summarizing vote counting between quantum centers, the initial quantum resources use two-particle product states, which are easy to prepare and require only single-particle measurements, making the operation convenient and reducing the difficulty of vote counting. This system can effectively detect and resist various attacks, thus ensuring security.

In the cell-free massive MIMO (CF-mMIMO) networks, characterized by differentiated service requirements, highly dynamic conditions, and decentralized resource deployment, the efficiency of distributing multi-dimensional network resources during CF-mMIMO caching deployment is constrained. To address this, this paper conducts research on the problem of diverse content caching and multi-user association in decentralized CF-mMIMO scenarios. First, based on the coupling relationship between content caching and user association, models for content caching, user association, and multi-dimensional resource allocation are studied and established. Second, given the stochastic and time-varying network environment and incomplete network state observations, the content caching, user association, and resource allocation problem are abstracted as a distributed partially observable Markov decision process (POMDP) with the objective of maximizing network efficiency. Taking into account the diverse content caching requirements and wide spatial differentiation, a multi-agent deep reinforcement learning algorithm based on graph attention network is further proposed for strategic learning and optimization of content caching, user association, and multi-dimensional resource allocation. Finally, simulation results confirm that the proposed algorithm significantly enhances performance in terms of network efficiency, system throughput and cache hit rate.

Open world object detection aims to simultaneously identify both known and unknown categories in dynamic environments, while enabling incremental learning of new categories. However, due to the lack of semantic representation ability of unknown categories, the guidance information between known and unknown categories is mutually coupled, resulting in limited detection performance. To solve this problem, this paper proposes an open world object detection based on causal prompt distillation, which innovatively combines visual-language model with causal inference to solve the problem of semantic bias between categories in open scenes. Specifically, by constructing a structural causal model, this paper reveals the semantic interference path between known and unknown categories from the perspective of causality. Then, causal prompt learning is proposed, which explicitly introduces the semantic prior of the open scene by generating semantic vectors of unknown categories to enhance the model's perception of unknown objects. Finally, in order to solve the problem of semantic bias in knowledge transfer, a causal distillation mechanism is proposed, and the guidance information of the known and unknown categories is decoupled by the double distillation loss decoupling teacher model. Experimental results demonstrate that this method has achieved good effects on multiple datasets, with an improvement in mean average precision (mAP) for known categories by 1.3% and a rise in recall rate (U-Recall) for unknown categories by 6.5%. These results validate the effectiveness and robustness of the proposed approach.

Multimodal intent recognition (MIR) is a critical research for understanding human intent in the real world. It aims to judge the speaker’s intent through multiple modalities including language, visual and audio modalities. However, existing studies in MIR primarily focus on constructing multimodal semantic environments for textual data, while the utilization of rich semantic information in visual and audio modalities, such as action and emotional semantics, remains insufficiently explored. Despite the visual and audio modalities carrying intents-related semantics, their inherent redundant information and noise hinder the effective use of these modalities. To address these challenges, this paper proposes a more effective MIR model that better leverages audio and visual information while suppressing redundant information. The proposed model understands the speaker’s intent by constructing primary semantic features that suppress redundant information and guiding the learning of intra-modality and inter-modality semantic associations at different scales. Based on this, the model leverages the potential intent consistency across different modalities and pair audio and visual representations with textual features, which contain more explicit intent-related semantics, to filter out irrelevant semantics that cannot be eliminated by intent recognition tasks. Furthermore, the model uses multi-modal fusion gating mechanism to integrate intent semantics from different modalities. Experiments on several datasets of intents understanding tasks show that the proposed method can effectively extract the modal semantics of audio and video and filter out the irrelevant semantics of intent recognition, and outperforms the existing MIR methods, achieving 0.7 to 1.8 percentage points improvement in accuracy (ACC), precision (P), recall (R) and F ₁ score (F ₁).

In recent years, autonomous driving has gained increasing attention due to its significant potential in improving road safety and enhancing traffic efficiency. The perception system plays a crucial role in modern autonomous driving systems, aiming to accurately estimate the surrounding environment's state and provide reliable observations for prediction and planning. Among them, 3D object detection serves as an important component of the perception system for predicting the positions, sizes, and categories of objects surrounding the autonomous vehicle. This paper provides a comprehensive overview of the research advancements in 3D object detection for autonomous driving in recent years. It discusses the advantages and limitations of single-modal methods and multi-modal fusion methods using different sensors from the perspectives of single-modal detection and multi-modal fusion detection. Furthermore, the paper compares the performance of various representative algorithms on public datasets, summarizes the current commonly used training strategies, and discusses the future development directions in this field.

The many-to-many communication routing problem is an NP-hard combinatorial optimization problem. Constructing efficient many-to-many communication routing paths requires timely acquisition of global network state information to adapt to the highly dynamic nature of network states. In this paper, within the context of software-defined wireless networks (SDWN), we address the issues present in existing data-driven multi-agent deep reinforcement learning methods, such as high computational and deployment costs, difficulty in adapting to the non-Euclidean characteristics of network topologies, excessive invalid actions during training leading to increased storage and time overheads, and slow convergence rates. This paper designs a new framework for collaborative sensing and intelligent decision-making between the SDN control plane and data plane and proposes a two-stage multi-agent routing method (Multi-Agent Graph deep reinforcement learning method based on intelligent node Deployment Strategy, MAGDS-M2M) to address the multi-to-multi communication routing problem. To reduce the computational and deployment costs associated with deploying agents on every node, a Q-learning-based intelligent node deployment algorithm is designed to determine the network nodes where agents need to be deployed. After completing the multi-agent deployment, a multi-to-multi routing decision method based on multi-agent graph reinforcement learning is developed within the Actor-Critic (AC) framework. This method redesigns the Actor and Critic networks using graph convolutional networks (GCN) and graph neural networks (GNN), addressing the weak adaptability of convolutional neural networks (CNN) to topological structure data in existing multi-agent reinforcement learning approaches. Additionally, to solve the issue of generating a large number of invalid actions during training caused by the fixed-length action space of the Actor network, a new local observation method for the action space is proposed. Experimental results demonstrate that the proposed method reduces task completion delay by 29.33% compared to benchmark experiments and verifies that by adjusting parameters, a balance can be achieved between task completion delay and the standard deviation of cumulative energy consumption across nodes. The source code developed in this work has been submitted to the open-source platform athttps://github.com/GuetYe/MAGDS-M2M.

The unmanned jammer is widely adopted in modern electronic warfare. However, it would reduce the signal quality of legitimate users while interfering with the enemy. To address this issue, an active jamming nulling strategy in the radio frequency domain is developed. By coordinating the waveform, amplitude, phase, and relative delay of signals emitted by dual unmanned jammers, a jamming nulling region is created while interfering with the enemy, ensuring the signal quality of legitimate users. Considering the inevitable time synchronization error between the dual unmanned jammers, closed-form expressions for the received signal-to-noise ratio and achievable rate advantages of legitimate users under the constraint of time error are given, which serve to assess the jamming nulling performance. Further, the emission power of unmanned jammers is optimized to maximize the achievable rate advantage for legitimate users, and the power optimization strategies are simplified according to practical jamming scenarios. Numerical simulations reveal that the proposed jamming nulling strategy outperforms traditional jamming strategies by an average of approximately 3.2 bps/Hz in achievable rate advantage. Compared with the jamming emission power strategy neglecting time synchronization errors, our proposed power optimization strategy has an average improvement of approximately 1.5 bps/Hz in achievable rate advantage.

Based on the low temperature co-fired ceramic (LTCC) three-dimensional packaging technology, this paper folds and vertically stacks multiple quarter-wavelength impedance transformation lines to achieve high integration in a multi-section broadband power divider. This design places 7 impedance transformation lines on the odd layers of the LTCC medium, and vertical vias are used to connect adjacent impedance transformation lines. Even layers are used to isolate the coupling effect between impedance transformation lines. The power divider not only achieves a relative bandwidth of 180%, but also has a size of only 4 mm×4 mm×1.33 mm. Comparing with the planar power divider with the same number of transmission sections, the horizontal size of this design is reduced by 84.6%. In the frequency range of 2~38 GHz, the measured values of S ₁₁ , S ₂₁ , S ₂₂ , S ₃₁ and S ₃₂ are better than - \mathrm{15} , - \mathrm{4.1} , - \mathrm{16} , - \mathrm{4.0} \mathrm{a} \mathrm{n} \mathrm{d} - \mathrm{17} dB, respectively. Since the power divider has the advantages of ultra-wideband characteristics, miniaturization and high integration, it can be widely used in mobile communications, radar detection, satellite navigation, industrial measurement and other fields.

Satellite signals in complex electromagnetic environments are often submerged in background and noise, and the performance of traditional signal detection algorithms degrades dramatically without accurate a priori knowledge. Currently, deep learning-based signal detection algorithms often require data post -processing steps that rely on expert experience and cannot achieve end-to-end detection of signals. To address the limitations of existing algorithms, an intelligent detection method of satellite signals based on DETR_S(DEtection with TRansformer on Signal) is proposed. Firstly, DETR_S is based on the coder-decoder architecture and uses the global modeling ability of the Transformer network to capture spectrum information. Secondly, uses the multi-head self-attention mechanism to effectively improve the problem of long-distance dependence of spectrum information. Then, the prediction frame matching module based on the Hungarian algorithm abandons the post-processing step of data with non-maximum suppression, and transforms the signal detection problem into a set prediction problem, so that the model can output the detection results in parallel. Finally, the signal reconstruction module is introduced, and the spectrum reconstruction loss function is added to the loss function to further improve the signal detection performance by mining the deep representation of the spectrum. The experimental results show that DETR_S can accurately detect faint satellite signals (>95%) at signal-to-noise ratios of 0 dB and above using only the signal spectral amplitude information, which is a significant improvement in the detection effect compared with the typical target detection representative network.

Non-contrast CT liver tumor imaging shows great potential in advancing the screening of colorectal cancer with liver metastasis. It provides reliable liver tumor segmentation from non-enhanced CT images, avoiding contrast agent toxicity, radiation, and costs. In this paper, we propose an innovative "teacher-student model driven by dual-modal knowledge collaboration(BKC-TS)"for accurately segmenting liver tumors in non-contrast CT images, significantly improving the safety, accuracy, and efficiency of liver tumor diagnosis and treatment. BKC-TS employs a teacher network to acquire explicit liver tumor knowledge and guide a student network in recognizing nearly invisible tumors from non-contrast images. It integrates clinical examination text data with medical imaging data. Text data, as prior information, guides tumor learning in CT images, enhancing precision and accuracy. (1) The Text-Image Collaborative Learning Teacher-Student Framework improves liver tumor segmentation accuracy in non-contrast images by integrating text knowledge and addressing CT image resolution issues. (2) The Dual-Modal Knowledge Fusion and Transmission Module combines imaging and clinical data through knowledge extraction, fusion, and transmission, effectively supporting tumor localization and recognition in non-contrast images. (3) The Gaussian Distribution-Constrained Student Self-Learning Strategy boosts the student network's independent learning, generalization, and robustness by iterating segmentation distribution and selecting beneficial knowledge. All experiments were conducted on a generalized dataset containing 1 140 CT liver images before and after enhancement. Experimental results show that BKC-TS achieved optimal liver tumor segmentation (at least a 2.17% IOU improvement), demonstrating its importance in non-contrast technology development.

Single-image high dynamic range(HDR) reconstruction can avoid ghosting artifacts that may be caused by multi-exposure HDR imaging, and is receiving widespread research. However, existing methods still struggle to effectively restore detail information in poorly exposed regions due to a lack of focus on critical information. To address this issue, this paper proposes a single-image HDR reconstruction method based on multi-attention and perceptual weighted learning, which aims to infer a high-fidelity HDR image from a single low dynamic range(LDR) image. Specifically, considering that the restoration of poorly exposed regions requires reference to compensation information from other regions, a multi-attention vision transformer(MA-ViT) with global-local receptive fields is designed. It combines depthwise separable convolution and attention mechanisms to achieve more effective global and local feature extraction and interaction. In addition, a loss aware weighted map is proposed to guide the network to focus on poorly exposed regions, further enhancing the quality of HDR reconstruction. Comprehensive comparative experiments are conducted on multiple benchmark datasets, and the results show that the proposed method improves the average peak signal to noise ratio(PSNR) by 0.23 dB compared to the state-of-the-art method, while generating HDR reconstruction results with higher visual quality.

Real conductors of interconnect structures are lossy and their skin depth becomes large at low frequencies. The traditional one-region formation with the approximation of Perfect Electric Conductor (PEC) or surface impedances may not be valid anymore, and two-region integral equation formations are needed in the integral equation approach. Also, the Electric Feld Integral Equation (EFIE) tend to break down at low frequencies and Augmented Electric Field Integral Equation(AEFIE) or Augmented Electric Field Integral Equations(AEFIEs) have been proposed to remedy the problem. In this work, we treat lossy conductors as penetrable objects and propose two-region Augmented Hybrid Field Integral Equations(AHFIEs) for low-frequency analysis. The Hybrid Field Integral Equations(HFIEs) consist of the EFIE of describing the exterior of a conductor and the Magnetic Field Integral Equation(MFIE) of describing its interior. Since the magnetic current density appears in the \mathrm{ℒ} operator in the HFIEs, we select the magnetic charge density as a new unknown function to be solved and introduce the continuity equation of magnetic current density as an extra equation. By incorporating the Volume Integral Equations(VIEs) of describing the substrate with arbitrary penetrable media in the interconnect structures, the two-region Augmented Volume-Surface Integral Equations(AVSIEs) are formulated for entire structures. The traditional method based on the AEFIEs can only be used for solving the problems including PEC interconnects and isotropic and homogeneous substrates while the proposed method based on the AVSIEs can applied to solve the problems with arbitrary materials so the capability of solving problems has been significantly enhanced. The AVSIEs are solved by the Method of Moments (MoM) where the Rao-Wilton-Glisson(RWG) and Schaubert-Wilton-Glisson(SWG) basis functions are used to represent the surface current densities of AHFIEs and volume current densities of VIEs, respectively, while a pulse basis function is employed to represent the charge densities of AHFIEs. Numerical examples are presented to illustrate the approach and good results have been obtained.

Mobile Crowd Sensing(MCS) collects data through the sensing devices carried by users and is a large-scale data sensing paradigm, where task allocation is one of the main challenges. This paper studies the task allocation problem of mixed users with heterogeneous quality delay-sensitive tasks. The design objective is to maximize the quality of task completion under the shared total budget of opportunistic users and participatory users. In response to the problem of insufficient prediction accuracy of existing prediction methods, this paper proposes a mobility prediction model based on transfer learning. By transferring the data of old participants with rich trajectories to new participants, it solves the prediction errors caused by the scarcity of historical data. Based on this prediction model, a mixed user task allocation algorithm is designed. The algorithm uses the mobility prediction model to allocate tasks to opportunistic users. In addition, the remaining tasks are clustered into different areas, and a bipartite graph matching problem is constructed to bind participatory users and task areas. Subsequently, an Ant Colony Optimization algorithm based on Travel Distance Balance(ACOTD) is proposed to achieve optimal path planning under the user's travel distance budget. Through a large number of simulation experiments on real datasets, this paper compares with existing algorithms. The results show that the algorithm has significant advantages in task completion quality and task allocation efficiency, verifying its effectiveness.

To address concept drift in non-stationary data streams that evolve over time, this paper proposes ICDC, an incremental density-based clustering algorithm designed for concept drift detection and adaptation over data stream. ICDC enhances the one-pass clustering framework by introducing a lazy outlier handling mechanism, where outlier evaluation is triggered by newly arrived data to distinguish between potential micro-clusters and noise. During clustering, data points and micro-clusters must satisfy feature dependency and temporal dependency conditions, effectively filtering outliers from the potential outlier set. This approach prevents irreversible deterioration of cluster structures caused by incorporating outliers—a limitation of existing outlier processing methods. Additionally, ICDC incorporates an outlier life cycle adjustment mechanism to control buffer size growth efficiently. By leveraging cluster structure changes as concept drift indicators, we propose a detection algorithm that enhances ICDC's sensitivity to local and global pattern shifts during data stream evolution. We evaluate ICDC on multiple real and synthetic dataset, assessing clustering quality, performance, concept drift detection and adaptation, memory overheade, and computational overhead. Experimental results demonstrate that ICDC outperforms existing algorithms on most datasets, achieving superior clustering accuracy and effectively detecting concept drift.

In image-level weakly supervised semantic segmentation(WSSS),class activation map(CAM) are commonly used to localize object regions. However, existing methods often encounter challenges such as under-activation in object regions and erroneous activation in background regions when generating CAM. This paper proposes a class-aware contrastive learning(CA-CL) framework for weakly supervised semantic segmentation, which significantly enhances the model's ability to accurately localize object regions by integrating text prompts and image category information. Firstly, we analyze the influence of different text prompt templates on the class activation maps of various categories, on this basis, to obtain more adaptive class representations, we construct a contextual prompt set and design a dynamic contextual prompt selection strategy. This strategy generates the most appropriate contextual prompts based on the similarity between image object regions and text prompts. Secondly, we adopt an image-text contrastive learning approach to enhance the model's performance in aligning image and text semantics, and we design a contrastive loss function to guide the model training process. Finally, we introduce a class-specific background suppression module to mitigate erroneous activation in background regions closely related to object categories, thereby generating more complete and compact class activation maps and achieving more precise semantic segmentation. Experiments conducted on benchmark datasets PASCAL VOC 2012 and MS COCO 2014 demonstrate the effectiveness of the proposed framework, achieving mIoU values of 71.9% and 43.9%, respectively. The results demonstrate superior performance compared to existing methods, significantly improving the accuracy of weakly supervised semantic segmentation.

To solve the problem of adapting the resource allocation in the data plane of polymorphic network to the network usage requirements of service traffic, this paper proposes a method for the coexistence and optimized deployment of network modals in a polymorphic network environment. This method considers key constraints such as polymorphic network element resource constraints, link resource constraints, and traffic forwarding delay to ensure the quality of user services. It also ensures the connectivity of each network mode through joint routing optimization. Since the direct modeling of the network modal deployment problem is a non-convex problem and difficult to solve directly, this paper transforms the original problem into a 0-1 integer linear programming problem. Based on this, a network modal deployment and routing selection algorithm based on stochastic rounding is designed. Theoretical analysis shows that this algorithm has an approximation ratio of {\rm O} ( \mathrm{l} \mathrm{o} \mathrm{g} n ), where n is the number of polymorphic network elements in the network. Simulation results demonstrate that the proposed approach can achieve efficient deployment of network modals in the data plane, effectively reducing link load by 13% to 22% while satisfying network resource and traffic forwarding delay constraints.

Based on the techniques of the support set exchange, the orbits of n k-variable k-rotation symmetric can be obtained from the orbits of n-variable 1-rotation symmetric through the analysis of the orbits of k-rotation symmetric.Furthermore, by modifying the support set of the rotation symmetric functions over {\mathbb{F}}_{\mathrm{2}}^{nk}, a new class of k-rotation symmetric 2-resilient Boolean functions are constructed and at least k \cdot {\mathrm{2}}^k different n k ( n \ge \mathrm{3} , k \ge \mathrm{2} )-variable k-rotation symmetric 2-resilient Boolean functions can be obtained.

To address the issues of a small quantity, large variability of real Web application attack data and diverse attack payloads that lead to poor training effects of large models, a network attack detection method based on federated large model (FL-LLMID) is proposed. Firstly, a federated learning network for fine-tuning large model is proposed. The server conducts incremental aggregation on the parameters generated by the client's local large model through incremental data training, which improves the parameter aggregation efficiency of large model in federated learning and avoids the problem of network traffic data exposure. Secondly, based on the large model ability to understand code, an attack detection model for application layer data (CodeBERT-LSTM) is proposed. By analyzing the application layer data packets, the CodeBERT model is used to perform vector encoding on the valid fields, and then combined with the Long Short-Term Memory Network (LSTM) for classification to achieve the attack detection task of Web applications. Finally, the experimental results show that the accuracy of the FL-LLMID method in the attack detection task for application layer data reaches 99.63%. Compared with traditional federated learning, the efficiency of incremental learning is improved by 12 percentage points.

Terahertz Synthetic Aperture Radar (SAR) exhibits broad application prospects due to its capability for high-resolution imaging and detailed target extraction. However, its short wavelength makes terahertz SAR extremely susceptible to the platform vibration, leading to many issues during the imaging process such as false imaging points, azimuthal blurring, and defocused SAR images. Therefore, this paper establishes a fine platform vibration model of terahertz SAR, and proposes an adaptive terahertz SAR motion compensation algorithm. Based on the impact mechanism analysis of the platform vibration on imaging using the mathematical model, the complex platform vibration in terahertz SAR imaging scenes can be compensated flexibly and accurately. Firstly, a fine terahertz SAR vibration model is established based on the Temporal Amplitude Modulation Vibration Model (TAMVM). By integrating the cosine time-varying amplitude and the random time-varying amplitude modulation vibration model, the TAMVM model reduces the limitation of the traditional harmonic model, and improves the adaptability to the complex and variable terahertz SAR platform vibration. Secondly, to address the performance loss of traditional harmonic model-based motion compensation algorithms when handling the complex platform vibration, this paper proposes an adaptive motion compensation method based on the Levenberg-Marquardt (LM) algorithm under the minimum Tsallis entropy criterion. The image quality-driven motion compensation algorithm proposed in this paper does not rely on the dominant target points, and it can precisely estimate the complex and varying vibration phase under the nonlinear least squares framework without the additional compensation steps. Moreover, the iterative process of the LM algorithm is derived under the minimum Tsallis entropy criterion in this paper. This algorithm adaptively adjusts the search displacement to achieve the feedback update and the iterative optimization, enabling precise estimation of the vibration phase and suppression of image blur, thereby obtaining high-quality focused terahertz SAR images. Furthermore, the comparison results of the simulated and real-measured data verify the rationality and feasibility of the proposed TAMVM model, and demonstrate the superiority of the proposed adaptive motion compensation method in achieving the precise terahertz SAR image focusing and suppressing false imaging points.

In recent years, with the widespread adoption of intelligent mobile devices and their powerful sensing capabilities, mobile crowdsensing (MCS) has emerged as a promising method for large-scale sensing of urban dynamics. A key challenge in MCS is discovering the truth from the noisy sensory data submitted by numerous workers. However, the process of truth discovery inevitably raises privacy concerns. To address these challenges, researchers frequently integrate local differential privacy (LDP) techniques by adding random noise to workers' data for privacy protection. Nonetheless, the randomness and unbounded nature of Laplace noise may inject excessive noise, resulting in outliers. Additionally, existing research often fails to adequately model the Laplace noise injected to satisfy LDP protection, resulting in low truth accuracy. Moreover, the current truth discovery methods are typically only applicable to discrete data, or cannot strictly satisfy the LDP constraints. To address the above issues, this paper proposes LEADER, an outlier-oriented truth discovery algorithm under LDP. First, the algorithm adds Laplace noise to workers' data to ensure privacy protection. Second, it addresses outliers by adopting the Huber loss function to measure distances, mitigating their impact on truth estimation. Finally, through a data-driven metric approach, the algorithm optimizes the weight allocation for worker and task importance and groups workers based on the similarity of their submitted values. These enhancements enable LEADER to improve the accuracy of estimated truths while maintaining privacy protection. Theoretical analysis demonstrates that LEADER strictly satisfies LDP constraints, effectively handles continuous data, and achieves high-accuracy truth discovery. Furthermore, compared to non-private truth discovery methods, the LEADER algorithm maintains comparable communication and computational overhead. Experimental results on two real-world datasets and a synthetic dataset indicate that the LEADER algorithm significantly outperforms existing methods, achieving at least an 18% improvement in the accuracy of the noisy truth.

To address the limited generalization capability of single capacity degradation models in predicting the remaining useful life(RUL)of lithium-ion batteries under varying operating conditions, this paper proposes a prediction method based on the interactive multiple model particle flow Filter(IMM-PFF). The method employs particle flow filter to collaboratively estimate the states of exponential, polynomial, and Verhulst models, and dynamically integrates multi-model predictions within an interactive multiple model framework, thereby adaptively matching the multi-phase characteristics of battery degradation. Experimental validation is conducted using lithium-ion battery degradation datasets(NASA and CALCE) under diverse operating conditions, which are divided into three distinct degradation phases. Results demonstrate that compared to single-model particle filter methods, the IMM-PFF reduces the root mean square error(RMSE)of capacity prediction and the absolute RUL prediction error by 24.3% and 4.5%, respectively. This study provides a novel high-precision and highly robust framework for lithium-ion batteries lifespan prediction in complex operational scenarios.

LiCi, LiCi-2, and GRANULE are all ultra-lightweight block ciphers designed for resource-constrained Internet of Things environments. Because of their fast encryption (or decryption) speed and favorable implementation in both hardware and software platforms, which have received extensive attention. In this paper, the linear structure characteristics of these ciphers are investigated via multiple perfect linear approximations (circular iterations) with an absolute correlation of 1. Moreover, the perfect linear approximations (linear distinguishers with probability one) for the full rounds of the LiCi, LiCi-2, and GRANULE are achieved, thereby completely breaking these cryptographic algorithms. It directly means that these block ciphers have serious design flaws.

Open-set malicious traffic recognition plays an important role in the field of network security. Existing methods have limitations in single model structure and lack of flexibility; neglecting incremental training samples selection, resulting in suboptimal classification performance. To address these problems, this paper proposes a method for continuous detection and classification of malicious network flows based on double-layer model and index distribution. Based on the relationship between the output weights of scalable extreme learning machine(S-ELM) and the standard output, this method designs following three indexes: the improved closest Pearson’s correlation coefficient, the normalized relative variance, and the normalized distance to "the others" column. These indexes are multiplied together to obtain a comprehensive index, which is combined with a single classifier for unknown class detection. In order to improve the continuous incremental capability of S-ELM in the open-set recognition task, a sample selection method based on the distribution of the comprehensive index is developed to select the optimal sub-dataset for incremental model training. Comparison experiments with existing representative methods show that the NA index of unknown class detection of the proposed method can be improved by 3%~13%, and the classification Acc index can be enhanced by about 3%~7% after continuous incremental updating.

With the rise of the internet of vehicles (IoV) and intelligent transportation systems, the increasing computational costs and problem scale have made the implementation of real-time applications extremely challenging, while also bringing a large number of combinatorial optimization problems that are in urgent need of parallel solving to vehicular edge computing (VEC). Often, these complex practical problems may possess non-convex, non-differentiable or even black-box objectives and constraints, which may go beyond the scope that traditional mathematical methods can handle. In this context, evolutionary multi-task optimization (EMTO), as a new paradigm in the field of multi-task optimization, effectively solves multiple independent optimization tasks in parallel by fully exploiting the potential correlations between tasks. An explicit EMTO framework tailored for IoV is designed. By integrating the unique characteristics of IoV tasks and deeply exploring the implicit correlations among them, a novel EMTO approach for IoV is proposed, which establishes mappings based on vehicle location information. This paper focuses on the multi-task optimization problem in the context of IoV, jointly optimizing fouraspects: vehicle-to-road data routing (DR), vehicle-to-road service migration (SM), vehicle-to-vehicle message transmission (MT), and vehicle-to-vehicle task offloading (TO), with the objective of maximizing the delivery rate of each task within a specified time frame. Furthermore, to enhance the efficiency of knowledge transfer among related tasks when their correlations are unknown, a migration mechanism grounded in task correlation assessment is introduced. Specifically, the longest common subsequence between links is utilized to calculate their similarity, and three migration strategies are devised according to different correlation distributions, ensuring the algorithm’s capability of knowledge transfer across various scenarios. Finally, through experimental validation and performance evaluation, the effectiveness of the proposed framework and algorithm is demonstrated. Compared with other EMTO algorithms, the algorithm presented in this paper exhibits faster convergence speeds for various optimization problems and yields better solutions after knowledge transfer among populations, showcasing impressive results.

High-precision direction of arrival (DOA) estimation is of great significance for multi-user high-speed communication using millimeter-wave large-scale arrays. To deal with the issues of degraded received signal quality due to wideband effects, reduced signal dimension caused by hybrid structures, and high computational complexity required in multi-user angle estimation, this paper proposes a high-precision DOA estimation method based on wideband signal phase measurement. Firstly, this paper establishes a system model and wideband received signal model for millimeter-wave large-scale arrays with a hybrid structure, and demonstrates the impact of wideband effects. Secondly, this paper derives the Cramer-Rao lower bound (CRLB) for DOA estimation and proposes an optimal training sequence design method that satisfies constant modulus constraints by minimizing CRLB. Subsequently, for single-user scenarios, this paper proposes a gridless high-precision DOA estimation method based on phase measurement reaching CRLB progressively. For multi-user scenarios, an iterative DOA estimation algorithm using the expectation maximization (EM) method is proposed on the basis of single-user DOA estimation to avoid dimension disaster caused by joint estimation and reduce computational complexity. Simulation results verify the effectiveness of the proposed algorithm. When the signal-to-noise ratio (SNR) exceeds 5 dB, the single-user and multi-user estimation algorithms proposed in this paper can progressively achieve CRLB, and the DOA estimation performance surpasses traditional estimation methods, avoiding the impact of wideband effects and signal dimension reduction.

In massive grant-free non-orthogonal multiple access (GF-NOMA) systems, multi-user detection usually relies on the prior sparsity of signals to detect active users. However, in practical applications, especially in dynamic multi-user access, the user access process becomes more complex and obtaining such prior information becomes more difficult. Therefore, this paper proposes a learnable threshold optimization scheme for massive dynamic multi-user access detection, namely the threshold-improved adaptive alternating direction method of multipliers (TI-A-ADMM) algorithm. In this algorithm, the time correlation of active user communication is utilized to introduce a dynamic correlation measure, which adaptively scales the noise threshold for active user detection, thereby improving detection performances. Moreover, to enhance the accuracy of active user detection across different signal-to-noise ratios, a deep learning network is employed to optimize the initial detection threshold, adapting to various access environments. Simulation results indicate that, in the case of dynamic multi-user access without known prior sparsity information, the proposed TI-A-ADMM algorithm achieves a performance gain of 2.4 dB in terms of active error rate (AER) and symbol error rate (SER) compared to existing algorithms with known sparsity information. The proposed algorithm exhibits lower performance degradation and higher robustness against interference caused by multi-user access.

To ensure the timeliness of multicast services in space-air-ground integrated networks, for minimizing energy consumption, this paper proposes an energy-efficient multicast routing algorithm based on multi-dimensional time-varying graphs. First, we analyze the energy consumption issues of multicast routing in low Earth orbit (LEO) satellite networks and develop a multi-dimensional time-varying graph model to characterize the time-varying topology, energy consumption, and delay of LEO satellite networks. Then, the K-shortest path (KSP) candidate path algorithm is applied to generate a path set that satisfies the quality of service (QoS) requirements of multicast services. A minimum path heuristic (MPH)-based multicast tree construction algorithm is further employed to derive the solution. Finally, it is theoretically proven that the multicast tree constructed by the proposed algorithm achieves near-optimal energy consumption under QoS constraints. Simulation results demonstrate that the proposed algorithm outperforms the existing methods in terms of network energy efficiency.

To address the issues of overlapping clusters, lack of spatial information consideration, and poor noise robustness in the possibilistic fuzzy C-Means (PFCM)algorithm, an enhanced PFCM algorithm integrating regional and neighborhood-level information is proposed. First, a novel function structure is designed to suppress overlapping clusters by introducing nonlinear attenuation characteristics, which effectively adjusts the contribution of different membership levels to various clusters, thereby reducing cluster overlap. Second, by incorporating local variance constraints, the algorithm integrates regional and neighborhood-level image information, fully utilizing spatial information to improve noise robustness. Finally, kernel metric is applied to the clustering dissimilarity measure, where the kernel bandwidth parameter is adaptively determined based on the intrinsic properties of the image, further enhancing algorithm flexibility. Segmentation experiments on noisy synthetic images, brain magnetic resonance imaging (MRI), and noisy color images demonstrate that the proposed algorithm achieves superior visual segmentation results and outperforms existing comparison algorithms in performance evaluation metrics.

With the rapid development of deep learning theory and technology, deep learning-based optical flow estimation methods have significantly improved in computational accuracy and robustness. However, due to the limitations of standard convolution's local receptive field and existing matching cost volume strategies that can lead to matching ambiguities, current methods often suffer from low accuracy in optical flow estimation and severe motion blur, particularly in large displacement motions and weak-texture regions. To address these issues, this paper proposes a global matching optimization optical flow estimation method combining deep separable residuals with multi-scale dual-channel attention. First, an encoding module is constructed that integrates deep separable residual blocks with multi-scale dual-channel attention, achieving more accurate depth features between consecutive frames while balancing parameter count and computational speed. Then, a learnable global matching optimization strategy for optical flow estimation is designed, which alleviates motion blur caused by matching ambiguities by excluding occlusions and efficiently utilizing global matching information. Finally, to enhance the model's training stability and generalization, a combined global and local optical flow loss function is proposed to constrain model training. Experiments conducted on the MPI-Sintel, KITTI-2015 and Middlebury test datasets demonstrate that the proposed method achieves the best optical flow estimation accuracy among all compared methods, especially showing better accuracy and robustness in large displacement and weak-texture regions.

This paper studies the stability and security of satellite network system to provide safe and reliable end-to-end service for business. Considering jamming attacks and denial of service (DoS) attacks in the satellite network, a joint detection security transmission scheme is designed according to the detection error accuracy to minimize end-to-end delay jitter. We construct the security transmission optimization problem based on the optimal control strategy, and determine the transmission path by combining the detection requirements as constraints. The optimal data transmission scheme is obtained by the augmented Lagrange differential evolution algorithm. For the first time, the accuracy of anomaly detection is used as a factor to determine the security path policy in this paper. When the network is attacked, the security control algorithm provides stable end-to-end services and the control policy dynamically triggers the anomaly detection, enabling the network to actively defend. In this paper, 66 satellite constellations are constructed to simulate and verify the effectiveness of the proposed secure transmission scheme in jamming attacks and denial of service attacks. The results show that using anomaly detection error as a decision factor of secure transmission strategy can effectively improve the stability of network services.

To address the issue of AMCL (Adaptive Monte Carlo Localization) failure in similar and dynamic environments within the field of mobile robotics, this paper proposes a method based on the improved YOLOv8 to construct a semantic chain list, which provides a pre-localization pose for AMCL, altering the particle weight update mechanism to enhance localization accuracy and robustness. Built on the YOLOv8 architecture, the method integrates the gather-and-distribute mechanism and attentional scale sequence fusion module to enhance the feature fusion capabilities of the Neck section, while pruning the model to improve both accuracy and speed. Laser SLAM is used to construct a 2D grid map, and the improved YOLOv8 extracts object semantics and maps them onto the grid map, generating a 2D semantic map. A semantic chain list is constructed based on the relationships between consecutive semantic objects. During localization, the robot's detected object semantic information is matched with the semantic chain list to provide a pre-localization pose for AMCL, modifying the particle update mechanism for precise localization. Additionally, a bag-of-words model is employed to mitigate semantic chain breaks caused by occlusion from obstacles. Localization experiments in similar and dynamic environments validate the effectiveness of the proposed algorithm.

In open-source software and platforms, developers can submit issues to report software bugs or suggest new feature requests. Due to the lack of experience and limited professional skills, users may struggle to summarize the content of issues accurately and effectively, resulting in low-quality issue titles, which in turn decreases the efficiency of addressing issues. Additionally, existing automatic issue title generation methods are primarily designed for English open-source platforms, such as GitHub, and the performance are degraded when applied to Chinese open-source platforms, like Gitee. Furthermore, existing methods mainly use the issue body description as inputs, ignoring the code snippets in the issue. In this paper, we propose a method called GITG (Gitee Issue Title Generation) specifically designed for Gitee, an open-source platform. GITG addresses the challenge of generating issue titles for both Chinese and English text by fine-tuning the Chinese BART pre-trained model on a constructed Gitee issue dataset. It leverages the bi-modal information from the issue body description and code snippets to generate informative and accurate issue titles. A dataset consisting of 18 242 Gitee issue samples is constructed to validate the effectiveness of GITG. Experimental results demonstrate that GITG outperforms iTAPE and iTiger by at least 13.09%, 10.18%, and 12.84% on the ROUGE-1, ROUGE-2, and ROUGE-L metrics, respectively. GITG also achieves improvements in BLEU and METEOR metrics. Human evaluation results also indicate that the average scores of the titles generated by GITG are improved by at least 26.7%, 20.8%, 24.2%, and 20.0% in overall score, fluency, informativeness, and conciseness, respectively, compared to iTAPE and iTiger.

To achieve customized communication and sensing service more flexibly, a beamforming optimization for coordinated rate-splitting multiple access (CoRSMA) assisted multi-static integrated sensing and communication (ISAC) is proposed in the paper. The received signal of common stream, private stream, and sensing echo from each base station (BS) is modeled, and the relationship between the beamforming vector and communication rate/sensing signal-to-noise ratio (SNR) is analyzed. Aiming to maximize the sum communication rate of the worst-case user equipment (UE) under the constraints of sensing performance requirement within a prescribed region, the beamforming vectors of all BSs are jointly optimized. To efficiently solve the formulated problem, the non-convex objective and constraints are relaxed into convex ones by introducing auxiliary variables and applying successive convex approximation (SCA) technique, and then the optimal beamforming vectors can be obtained via semi-definite programming (SDP) technique. Simulation results demonstrate that our proposed CoRSMA-ISAC system outperforms the ISAC system with spatial division multiple access (SDMA) and non-orthogonal multiple access (NOMA) in terms of both communication and sensing performance.