Infrared image generation is essential in scenarios with low illumination or complex environments, but the scarcity of aligned visible–infrared data and the lack of physical realism in generated results remain key challenges. However, existing generative models often overlook the thermodynamic principles underlying infrared imaging, resulting in synthetic images that are visually plausible yet physically inaccurate. In this paper, we propose Infrared Physics-guided Latent Diffusion (IPLD), a novel framework that integrates physics-guided modeling into a latent diffusion process for high-fidelity synthesis of infrared images. Central to IPLD is the Temperature–Emissivity–Environmental Radiance (TeR) decomposition, which decomposes thermal signals into temperature, emissivity, and environmental radiance components, governed by the laws of blackbody radiation. To enhance the environmental radiance modeling, we introduce Environmental Radiance Map Estimation (ERME), a hybrid local–global estimation mechanism that preserves both spatial detail and thermal consistency. Furthermore, a novel Skip Connection Diffusion Transformer (SCDT) is proposed to strengthen and balance semantic structure and fine-grained details during image reconstruction. Extensive experiments on public datasets demonstrate that IPLD outperforms state-of-the-art Generative Adversarial Network (GAN)-based and diffusion-based methods, achieving superior results in Structural Similarity Index Measure (SSIM), Peak Signal-to-Noise Ratio (PSNR), Learned Perceptual Image Patch Similarity (LPIPS), and Fréchet Inception Distance (FID) metrics. Ablation studies validate the complementary value of TeR, ERME, and SCDT in improving radiative realism. Our approach establishes a new paradigm for physically grounded image translation, offering enhanced generalization and reliability for downstream perception tasks such as target detection.

Nowadays, short-term time-series prediction is increasingly vital across domains such as transportation, energy, and weather forecasting. Its high accuracy and rapid response provide crucial support for dynamic changes in these applications. Transformer-based methods, with their powerful temporal feature extraction, have emerged as a leading paradigm for this task. However, many Transformer-based methods exhibit excessive complexity for short-term horizons and rely too heavily on long-term dependencies. In this article, we propose the specialized sparse Transformer without shortcuts (SSTwS) to approach short-term multivariate time-series prediction (SMTP). We specialize in the Transformer-based prediction methods from two perspectives: simplifying the model structure and improving the self-attention mechanism. The SSTwS aims to balance performance and cost. Specifically, we removed residual connections designed for deep networks and enhanced convergence speed through the prepositioned layer normalization (LN). We developed a specialized sparse-centering self-attention mechanism, removed half of the linear projections typical in vanilla self-attention mechanisms. In addition, we leveraged the sparsity of the self-attention mechanism, employing five probabilistic methods combined with centering to select the dominant features. Extensive experiments on seven widely used datasets demonstrate the superiority of our SSTwS in terms of training speed and performance relative to existing state-of-the-art methods.

Accurate detection of small object clusters in modern security surveillance systems is critical for threat prevention and response, directly impacting monitoring efficiency and risk management efficacy. However, prevailing object detection algorithms struggle with high-density small target scenarios, suffering from slow inference speeds, insufficient precision, frequent false positives, and high miss rates. These limitations undermine real-time reliability and responsiveness to emergencies. To address these challenges, we propose SIEP-YOLO, an enhanced YOLOv11-based model integrating three novel components: the SDI-iAFF feature fusion module, EUCB upsampling module, and a specialized small object detection layer P2. This architecture boosts representational capacity and detection performance while simplifying computational complexity, achieving a balance between high accuracy and lightweight design. Experimental results demonstrate that SIEP-YOLO outperforms the original YOLOv11 by 3.2% and 2.0% in [email protected] and mAP@(0.5:0.95), respectively, on benchmark datasets. The proposed model thus emerges as a superior solution for small object cluster detection, enabling more efficient and reliable surveillance in complex environments.

Military aircraft target detection remains a critical challenge in modern defense systems, particularly for remote sensing imagery with complex environmental interference. Existing approaches often exhibit limitations in maintaining high detection fidelity across varying resolutions and cluttered backgrounds. To overcome these constraints, this study presents SR-YOLOv11, a multi-stage framework integrating Super-Resolution (SR) reconstruction and hierarchical feature optimization. Initially, the Enhanced Deep Super-Resolution (EDSR) network is deployed to refine input image quality, ensuring precise preservation of critical aircraft morphological features. Subsequently, the YOLOv11 architecture is systematically enhanced through three key innovations: 1) Replacement of the native C3k2 module with a C3k2_AdditiveBlock to amplify discriminative feature learning; 2) Integration of an Adaptive Downsampling (ADown) layer for computationally efficient multi-scale context aggregation; 3) Implementation of an Auxiliary detection (AuxDetect) head mechanism with cross-layer feature fusion, significantly boosting localization accuracy for occluded targets. Comprehensive evaluations on the MAR20 dataset demonstrate the framework’s superiority, achieving 98.7% mAP@50 and 80.9% mAP@(50:95). The proposed architecture demonstrates enhanced robustness in complex environments while maintaining real-time processing efficiency, validating its operational viability in aerial surveillance scenarios.

Focusing on the initial alignment problem of inertial navigation system (INS) when GNSS-denied, we propose a deep learning inertial navigation initial alignment framework (IANet), construct an INS initial alignment model when GNSS-denied, and design a data-driven velocity estimator. According to INS principle, a X Temporal Convolutional Networks (XTCN) is proposed, which takes inertial device data as network input and velocity increment as network output. XTCN fully exploits the ability of feature extraction and information association in TCN residual module, deeply explores the coupling relationship between inertial device data and velocity increment, and effectively improves velocity estimation accuracy. IANet provides effective parameter information reference for INS initial alignment when GNSS-denied, which improves initial alignment accuracy. Simulation and experiment results show that XTCN can not only ensure that INS can independently and effectively complete the initial alignment when GNSS-denied, but also improve the yaw angle alignment accuracy by 56%.

In urban safety, intelligent transportation, and smart security applications, robust pedestrian detection is paramount. Methods that rely solely on visible light imaging struggle in low-light or adverse weather conditions. To address these challenges, we propose dual-modality pedestrian detection (DMPD)—a novel dual-modality pedestrian detection framework that fuses visible and infrared imaging through innovative fusion strategies. The method integrates a modal alignment module to reduce pixel-level misalignment, a differential modal fusion module to effectively combine complementary features while suppressing noise, and a mix module that enhances multiscale feature extraction via integrated convolution and self-attention mechanisms. Furthermore, the enhanced YOLOv7 is used to further boost feature representation and detection accuracy. Experimental results on the public dataset demonstrate that DMPD achieves a detection$mA{{P}_{50}}$of 97.1% and a real-time speed of 118 FPS, outperforming state-of-the-art methods under both normal and adverse conditions, including fog, rain, and snow. These results confirm the effectiveness of the proposed fusion strategy in harnessing the complementary strengths of visible and infrared modalities, thereby offering a highly robust and scalable solution for pedestrian detection in complex urban environments.

Infrared small target detection (ISTD) holds significant importance in domains such as maritime search and rescue, and autonomous driving. To enhance the detection capabilities of infrared small targets against complex backgrounds, a novel detection algorithm based on an improved Detection Transformer (DTER) is proposed. This algorithm leverages the DTER detection framework and the EDSR network, utilizing super-resolution reconstructed images as inputs. It incorporates the Enhanced Multi-Scale Attention (EMA) module and an improved backbone structure. Moreover, it employs a micro-target detection encoder head with a new feature layer S2 to elevate the quality of minute feature extraction. The proposed method achieved a mAP@50 of 96% and mAP@(50:95) of 54.6% on a public dataset. Compared to current state-of-the-art methods for infrared small target detection, it demonstrates superior capabilities in reducing false positives and misses while maintaining commendable real-time performance.

Semantic segmentation in nighttime scenes presents a significant challenge in autonomous driving. Unsupervised domain adaptation (UDA) offers an effective solution by learning domain-invariant features to transfer models from the source domain (daytime scenes) to the target domain (nighttime scenes). Many methods introduce a latent domain to reduce the difficulty of UDA. However, they often build only a single adaptation pair of “latent-to-target”, which limits the effectiveness of knowledge transfer across different domains. In this letter, we propose a Masked Guided Meta-Learning (MGML) framework for domain-adaptive nighttime semantic segmentation. Within the MGML framework, we explore two key issues: how to generate the latent domain, and how to leverage the latent domain to assist meta-learning in reducing domain discrepancy. For the first issue, we employ the fast Fourier transform along with a complementary masking strategy to generate masked latent images that resemble the target scenes in the latent domain without adding to the training burden. For the second issue, we nest a mask-based consistency constraint within a bi-level meta-learning framework, enabling cross-domain knowledge acquired from the pair of “source-to-latent” to enhance the “latent-to-target” adaptation. Experiments on benchmark datasets demonstrate that our MGML achieves state-of-the-art performance, demonstrating the effectiveness of our approach in nighttime semantic segmentation.

The recent development of advanced trackers, which use nighttime image enhancement technology, has led to marked advances in the performance of visual tracking at night. However, the images recovered by currently available enhancement methods still have some weaknesses, such as blurred target details and obvious image noise. To this end, we propose a novel method for learning multidimensional spatial attention for robust nighttime visual tracking, which is developed over a spatial channel transformer based low light enhancer (SCT), named MSA-SCT. First, a novel multidimensional spatial attention (MSA) is designed. Additional reliable feature responses are generated by aggregating channel and multi-scale spatial information, thus making the model more adaptable to illumination conditions and noise levels in different regions of the image. Second, with optimized skip connections, the effects of redundant information and noise can be limited, which is more useful for the propagation of fine detail features in nighttime images from low to high level features and improves the enhancement effect. Finally, the tracker with enhancers was tested on multiple tracking benchmarks to fully demonstrate the effectiveness and superiority of MSA-SCT.

Recent advanced trackers, composed of discriminative classification and dedicated bounding box estimation, have achieved remarkable advancements in performance of visual object tracking. However, existing methods cannot satisfy the demands of tracking tasks in complex scenes, such as occlusion, scale variations, and etc. To this end, we propose a novel online multi-scale classification and global feature modulation for robust visual tracking, which is developed over accurate tracking by overlap maximization, named ATOM+. First, coordinate attention (CA) is applied to enhance the target features in the channel dimension and spatial dimension, which can effectively optimize the feature representation ability of the backbone network. Second, an online multi-scale classification (OMC) module is designed. During the online tracking phase, more reliable matching responses are comprehensively generated by aggregating information from different scales related to the target. This new operation enables stable perception of the target by the tracker, particularly when severe changes in the appearance and posture of the target are encountered. Third, a global feature modulation (GFM) mechanism is constructed, which requires only a small amount of computational resources, to fuse the spatial contextual information of the template image into the search region. This integration refines the bounding box to obtain an accurate estimate of the target state. Finally, comprehensive experiments on conventional tracking benchmarks of OTB100, LaSOT, and VOT2018 show that our tracker can sufficiently address different challenging scenarios, and achieves state-of-the-art performance. For the average running speed, our tracker can achieve 37 FPS in real time.

Position and location constitute critical context for Internet of Things (IoT) devices. Global navigation satellite systems (GNSSs) are the primary apparatus providing precise position and location information for IoT devices in outdoor environments. However, in dense urban areas, non-line-of-sight (NLOS) signals will induce large errors in GNSS pseudorange measurements due to the additional signal transmission paths. The vector tracking (VT) technique utilizing a Kalman filter (KF) to estimate navigation solutions has been investigated in NLOS detection, and its advantages have been demonstrated. However, the estimation of NLOS-induced bias has not been thoroughly investigated in the VT framework. In this article, we focus on the estimation and correction of NLOS-induced errors within the VT framework. First, graph optimization (GO) instead of a KF is incorporated with VT to optimize the estimation of navigation solutions. The NLOS-induced bias is then added to the VT state vector as the variable for real-time estimation. Compared with the KF-VT method, in GO-VT, the state transformation and the measurement model are regarded as constraints to optimize the state vector estimation. Hence, the GO-VT framework is more flexible than the KF approach in dealing with state vector changes. An iterative process is conducted to solve for the optimization results; a multiple-correlator scheme is employed in GO-VT to provide the initial values of the NLOS-induced bias. Three collected GPS L1 data sets (static and dynamic) are used to evaluate the proposed method. The statistical results support the conclusion that GO-VT with state augmentation achieves superior position estimation in urban areas.

This article focuses on the consensus control problem for a class of time-varying stochastic nonlinear time-delay multiagent systems (MASs) attacked by deception attacks. The stochastic deception attack is considered in the procedure of propagating measurement information among agents. To solve the consensus control problem for addressed MASs under stochastic deception attacks, a definition of quasi-consensus is put forward. The objective of our investigation is to devise a consensus protocol to drive all agents to stay within an allowable range despite the existence of stochasticity and external malicious attacks. With the help of recursive linear matrix inequality and stochastic analysis methods, sufficient conditions are acquired to guarantee that all agents are constrained in the desirable range. Subsequently, an optimization algorithm is presented, which is to seek the locally optimal allowable distance among agents. Finally, a simulation example is presented to demonstrate the availability of our proposed algorithm.

Internet of Vehicles (IoV) is a promising Internet of Things application, where roadside unit (RSU) plays an important role for network service provisioning. How to select the number and locations of RSUs to deploy and allocate the traffic load to them is a critical and practical open problem. Most of the existing work focused on 1-D scenarios assuming unlimited RSU capacity, while a more practical 2-D case with limited RSU capacity has not been fully considered yet. In this paper, we investigate an RSU deployment problem for 2-D IoV networks considering the expected delivery delay requirements and task assignment. We formulate a novel utility-based maximization problem to solve the RSU deployment problem, where the utility function indicates the total benefit from the RSU deployment. We observe that each RSU has an irregular service area, which makes the problem much more difficult than the traditional facility location problem. Then, we design a utility-based RSU deployment algorithm (URDA), a linear programming-based clustering algorithm, to solve the problem. The gap between URDA and the optimal solution has been analyzed, which proved that the proposed URDA is near optimal if the deployment cost is low. Extensive simulations have been conducted to demonstrate the effectiveness and superiority of the proposed solution for IoV network service guarantee over other approaches.

Abstract Deviation registration algorithm of homogeneous sensors has grown mature, but heterogeneous sensors deviation registration algorithm still faces the problems like strong nonlinearity and low precision, which needs urgent solutions. To eliminate the difficulties in deviation registration of heterogeneous sensors, especially the different dimensional sensors, the heterogeneous sensors deviation registration model and dynamic adaptive particle swarm optimized particle filter are proposed. In this paper, dynamic adaptive particle swarm optimized particle filter and Kalman filter is federated to carry out the real‐time calibration for the system deviation of radar and infrared sensor. With particle neighborhood information taken into account, this algorithm conducts a self‐adaptive adjustment over the number of particle's neighborhood particles by means of diversity factor, neighborhood extension factor, and neighborhood restriction factor. The adjustment is intended to control the influence of particle on neighborhood and reduce its local optimum so as to attain the optimal balance between convergence rate and optimization ability. The experimental result shows that the improved heterogeneous sensors deviation registration algorithm has upgraded the accuracy and speed of deviation registration; therefore, this algorithm is of high application value in the deviation registration of three‐dimensional radar and two‐dimensional infrared sensor.

Abstract Track‐before‐detect algorithm based on the particle filter algorithm has the problems of low tracking precision, poor particles, and requiring a large amount of particles to be calculated in a low signal‐to‐noise ratio, which is difficult to meet the accuracy and speed required by the modern infrared search and tracking system. In this paper, an improved infrared small target detection and tracking method based on a new particle filter is proposed. This is where particles are used to represent an individual bat to imitate the hunting process of bats. By adjusting loudness, frequency, and impulse emissivity of a particle swarm, the optimal particle at that time is followed to search in the solution space. In addition, the global search and the local search can also be dynamically switched to improve the quality and distribution of the particle swarm. The performance of the proposed algorithm is tested in a simulation scene and the real scene of the infrared small target detection and tracking. Experimental results show that the proposed algorithm improves the performance of the infrared searching and tracking system.

Recently, a major effort has been made to develop methods or tools for rock characterization and mineral content mapping. Light detection and ranging (LiDAR) is an efficient active remote sensing technique for collecting geometry information about rock surfaces. However, traditional LiDAR sensors work with a single-wavelength laser source, and it is unfeasible to obtain spectral information using one LiDAR sensor. The combination of hyperspectral imaging and LiDAR techniques is an emerging method for acquiring spatial and spectral information simultaneously that allows remote mapping of high-resolution mineral content and distributions and identifies subtle chemical variations. Unfortunately, spatial and spectral data registration, which introduces additional complicated data processing, is an inevitable and essential issue for this method. In this letter, first, we investigate the feasibility of ore classification applications with hyperspectral LiDAR (HSL). HSL consists of 17 spectral channels covering the visible–shortwave infrared (SWIR) spectral range. Spatial and spectral information about seven different ore samples is obtained under a controlled laboratory environment using HSL. The standard deviation of the distance measurements is less than 1.1 cm for different spectral channels, and the classification accuracy can reach 100% if all 17 spectral measurements are used. To optimize the system design with lower cost and system complexity, a spectral band selection criterion is built based on the feature contribution degree (FCD), which is calculated using the normalized variance of the reflectance values for different ore samples at each wavelength. Two different strategies of FCD selection are tested to generate vectors: ascending sequences and descending sequences. Feature vectors with descending sequences have better classification accuracy. In addition, the results show that the classification accuracy can reach 100% with the feature vector of the seven largest FCD values compared to 59.57% for the feature vector with the seven smallest FCD values. Moreover, we find that the channels with high FCD values are primarily centered in SWIR bands. This result could be a reference for optimizing the hardware design of HSL for ore classification or mineral identification.

To meet the requirements of modern radar maneuvering target tracking system and remedy the defects of interacting multiple model based on particle filter, noninteracting multiple model (NIMM) and enhanced particle swarm optimized particle filter (EPSO‐PF) are proposed. The improved maneuvering target tracking algorithm (NIMM‐EPSO‐PF) in this article combines the advantages of NIMM with those of EPSO‐PF. NIMM is used to figure out the index of particles to avoid the high computing complexity resulting from particle interaction, and EPSO‐PF can not only improve the equation of particle update through the rules individuals develop an understanding of group but also enhance particle diversity and accuracy of particle filter through the small variation probability of superior velocity. Besides, the random assignment of inferior velocity is capable of upgrading filter efficiency. As shown by the experimental result, the NIMM‐EPSO‐PF not only improves target tracking accuracy but also maintains high real‐time performance. Therefore, the improved algorithm can be applied to modern radar maneuvering target tracking field efficiently.

Future hybrid vehicle networks can use both Vehicle-to- Infrastructure (V2I) and Vehicle-to-Vehicle (V2V) communications to provide reliable, timely, scalable, and media-rich services. In this paper, we investigate the problem that how to disseminate the data to the Road Side Unit (RSU) considering bidirectional transmissions, using vehicles to store-carry-and-forward the messages if possible, in hybrid V2I/V2V networks. We focus on the delay modeling and dissemination strategy design, aiming to minimize the delivery delay. Considering a one-dimensional vehicle network with multiple road segments, we model the process of uploading a message to an RSU either in front of or behind the source. Furthermore, based on the delay analysis, we obtain the desirable message dissemination direction, and further design the message uploading algorithm to minimize the expected deliver delay. Simulations have been conducted to verify the correctness of the analysis and illustrate the efficiency of the proposed algorithm. The analytical model can also provide important insights and guideline for the deployment of RSUs.

A novel near-space floating platform aided passive location method for super-low-altitude target using time difference of arrival(TDOA) is proposed. The new method uses a near-space floating platform as the baseline in the height to form a redundant location system consisting of several subsystems. Each subsystem can obtain a set of location results with ambiguity. After eliminating the ambiguity by the nearest matching, the new method can improve the location accuracy in the whole observed airspace by simplified weighted least square(SWLS) fusion. A better tracking performance can be achieved by treating the SWLS fusion location result as pseudo-linear measurement, and keeping the tracking through Kalman filter. Simulation results show that the proposed method comes to stabilization within 40s and has higher tracking precision than the SWLS method.