Feature point detection on textureless surfaces remains a fundamental challenge in computer vision due to the absence of discernible color and brightness gradients. From the imaging mechanism perspective, micro-geometry structures of textureless surfaces provide physically stable cues for feature point extraction despite the absence of visual distinctiveness. Therefore, we propose a novel feature point detection method, which reconstructs surface micro-geometry structures from a single RGB image and leverages these micro-geometry structures for feature extraction, without relying on specialized equipment or complex deep learning models. Specifically, our method establishes a novel framework that models the light-surface interaction to analyze phase modulation in reflected light. Then it reconstructs underlying micro-geometry structures through Gabor Kernel-based spectral analysis, enabling accurate quantification of surface height variations from phase information. This information forms the foundation of our proposed Concave-Convex Index (CCI), a robust geometric descriptor that achieves stable feature characterization through geometry-aware measurements. Extensive evaluations on TUM, T-LESS, Shape2.5D datasets and self-collected images, demonstrate our method's superior capability in extracting stably distributed and highly repeatable feature points, even when visible texture or brightness gradients vanish. Our method offers a novel perspective for reliable feature point detection on challenging textureless surfaces across diverse materials and illumination conditions.

Surface normal estimation is a fundamental task in point cloud processing and plays a crucial role in downstream applications. Existing methods typically extract features from local neighborhoods or patches, followed by surface fitting or direct regression to predict normals. However, the scale ambiguity in determining the optimal neighborhood hinders effective extraction of geometric information, making normal estimation for unstructured point clouds with significant density variations particularly challenging. To address this challenge, we propose APN-Net, an adaptive perception network for point cloud normal estimation. Specifically, we design the Graphical Information Self-perception (GIS) module, which provides an implicit manner for region partitioning and expands the receptive field, enabling automatic extraction of both local geometric details and global structural information, while alleviating the scale ambiguity in determining the optimal neighborhood. Moreover, to capture complex geometric details, we introduce the Adaptive Graph Convolution (AGC) module, which employs adaptive kernels to model relationships among points across different semantic regions, thereby enabling richer feature representation. Extensive experiments on both synthetic and real-world scanned datasets demonstrate that APN-Net achieves superior performance in unoriented normal estimation, particularly for point clouds with significant density variations.

Accurate sleep staging is essential for diagnosing sleep disorders and improving sleep health. While traditional methods rely on multichannel electroencephalogram (EEG) signals, single-channel EEG offers a more practical and non-intrusive alternative. However, the complexity of sleep dynamics across varying time scales presents a significant challenge for single-channel EEG-based staging. To address this, we propose a novel method combining Multi-Scale Attention and Cross-Time Learning to capture both local and global temporal dependencies in EEG signals. The Multi-Scale Attention Module extracts features at multiple temporal scales, while the Cross-Time Learning mechanism models long-term dependencies across time segments, improving classification performance. Our approach is evaluated on benchmark EEG sleep datasets, demonstrating superior accuracy and F1-scores, particularly in transition stages, compared to existing single-channel methods. The results suggest our method’s potential for real-world applications in portable and home-based sleep monitoring systems.

Automatic sleep staging is crucial for sleep assessment and diagnosis. Signals of different modalities, such as electroencephalogram (EEG) and electrooculogram (EOG), are of crucial importance for sleep staging. Therefore, effective fusion of different modal signals is the key to improve sleep staging performance. Current methods usually ignore the inter-modal interactions in the process of multimodal signals fusion, making it difficult to achieve optimal sleep staging performance. In order to solve this problem, we propose a sleep staging model, referred to as CASleepNet, which can extract the heterogeneous features of each modality through Temporal Context Module (TCM), and learn the intrinsic connection between different modal signals by using Dual Cross Attention Fusion (DCAF). Experimental results show that CASleepNet can effectively fuse multimodal information to improve the performance of sleep staging.

The joint classification of hyperspectral image (HSI) and light detection and ranging (LiDAR) data seeks to provide a more comprehensive characterization of target objects. Multi-modal data possess distinct semantic structures in both spectral and spatial dimensions, making efficient feature complementarity and redundancy elimination crucial. To this end, we propose a self-Distillation-based multimodal Feature Alignment Network (DFANet), which employs two branches to capture spectral and spatial similarities respectively, and integrates structural discriminative information from LiDAR at two stages for more effective multimodal data integration. The network comprises three main components: a Feature Alignment Fusion Module (FAFM), an Offset Attention Module (OAM), and a self-distillation mechanism. Specifically, the FAFM guides feature alignment through channel-assimilative mapping of multimodal data. The OAM addresses boundary patch classification challenges by learning offset weights of reference points. The self-distillation mechanism filters out irrelevant information during feature alignment by enhancing the coordination between high-level and low-level features. Adequate experiments indicate that our method achieves better results compared to the most recent hyperspectral classification methods on three public datasets.

Remote sensing (RS) images have rich ground information, diverse object types, and large-scale differences, and these characteristics make difficulties in achieving precise segmentation. Recently, the state-space model improved by Mamba offers global modeling capability while maintaining linear computational complexity. However, it still faces issues with insufficient extraction of global information specific to the spatial and channel dimensions, which is crucial for achieving accurate segmentation, along with lacking sensitivity to local details. Based on this, we propose a hybrid Mamba-attention fusion network (HMAFNet) for RS image semantic segmentation, based on the encoder-decoder architecture. Specifically, the encoder incorporates the spatial-channel Mamba (SCMamba) module, which uses the Mamba to efficiently capture global feature representations across both spatial and channel dimensions. Meanwhile, local information essential for the encoding phase is supplemented by a parallel convolutional branch. In the decoding phase, we propose the information-guided cross fusion (IGCF) module, which generates corresponding features via convolution-based and Mamba-based information-guided branches. The cross-attention mechanism facilitates the interaction and fusion between the features, thereby preserving elaborated details and further eliminating semantic differences. Extensive comparison experiments and ablation experiments on both the Vaihingen and Potsdam datasets show that our proposed HAMFNet can achieve better segmentation results.

Spatio-temporal action detection networks, which need to simultaneously extract and fuse spatial and temporal features, often result in existing models becoming bloated and difficult to run in real-time and deploy on edge devices. This paper introduces an efficient and real-time spatio-temporal action detection model, YOWOv3. This model uses efficient 3D and 2D backbone networks to separately extract spatial and spatial-temporal features from sequential information. A lightweight spatio-temporal feature fusion module, designed by deeply integrating convolution and self-attention mechanisms, further enhances the extraction of spatio-temporal features. We refer to this module as the CFACM (Channel Fusion & Attention Convolution Mix) module. Our approach not only outperforms the latest efficient spatio-temporal action detection models in terms of lightness, reducing the model size by 24% compared to the latter, but also improves the mAP accuracy on the UCF101-24 dataset by 1.35%, while maintaining excellent speed performance, thus achieving a balance between accuracy and speed. Furthermore, existing models often use 3D convolutions to extract temporal information, which may be limited on certain devices, such as Apple’s M series processors. To mitigate the potential issue of 3D convolution operators not being supported during edge deployment of spatio-temporal action detection models, we employ a spatio-temporal shift module containing only 2D convolutions. This enables the model to acquire temporal information and inject the obtained temporal features into multi-level spatio-temporal feature extraction models. This not only liberates the model from the constraints of 3D convolution operations but also enhances the model’s balance between accuracy and speed. This results in state-of-the-art performance in lightweight networks using only 2D convolutions.

Belief rule base (BRB) has attracted considerable interest due to its interpretability and exceptional modeling accuracy. Generally, BRB construction relies on prior knowledge or historical data. The limitations of knowledge constrain the knowledge-based BRB and are unsuitable for use in large-scale rule bases. Data-driven techniques excel at extracting model parameters from data, thus significantly improving the accuracy of BRB. However, the previous data-based BRBs neglected the study of interpretability, and some still depend on prior knowledge or introduce additional parameters. All these factors make the BRB highly problem-specific and limit its broad applicability. To address these problems, a model-agnostic statistical BRB (MAS-BRB) modeling approach is proposed in this article. It adopts an MAS methodology for parameter extraction, ensuring that the parameters both fulfill their intended roles within the BRB framework and accurately represent complex, nonlinear data relationships. A comprehensive interpretability analysis of MAS-BRB components further confirms their compliance with established BRB interpretability standards. Experiments conducted on multiple public datasets demonstrate that MAS-BRB not only achieves improved modeling performance but also shows greater effectiveness compared to existing rule-based and traditional machine learning models.

Balancing image local detail enhancement with brightness enhancement has been a challenge. The images captured by wireless capsule endoscopy (WCE) are low-light and unclear. To this end, we propose an adaptive enhancement method for WCE images. Firstly, we use the guided filter to filter and smooth the WCE images to approximate its illumination component, and then the reflection component is obtained by decomposing it based on the Retinex model. Secondly, an adaptive Sigmoid function is obtained according to the positive correlation between the just-noticeable difference (JND) threshold of the illumination component and the gain parameter of the Sigmoid function, to adaptively enhance the illumination component, and then based on the Retinex model fusion with the reflection component. Finally, we combine with Gamma correction algorithm to enhance the contrast of the above results. Experimental results show that the proposed method can adaptively enhance the overall effect and local details of WCE images; and the feature extraction and matching effects are better than the classical enhancement algorithms, with an average increase of 88.8% and 59.1% respectively.

In this paper, we propose an improved distributed fusion algorithm under the Labeled multi-Bernoulli (LMB) filter framework. Firstly, the LMB parameter set is augmented by a new group variable, which is able to record the matching information of the neighbour sensors. Then the matching LMB components of the survival targets between the sensors can be fused directly by checking the group variable, greatly reducing the time cost of the matching calculation and the interference from the newborn targets. While for the newborn targets, the Murty algorithm is employed and only performed once to find the best matching relation between the sensors. Finally, experimental results show that the proposed algorithm offers a better tracking performance than the state-of-the-art R-GCI-LMB algorithm with lower computational complexity and higher tracking accuracy.

Significant progress has been achieved in remote sensing image scene classification (RSISC) with the development of convolutional neural networks (CNNs) and vision transformers (ViT). However, high intra-class diversity and inter-class similarity are still enormous challenges for RSISC. Metric learning can effectively improve the discriminative ability of deep representations by constraining the distance between features. Previous metric learning methods only optimize the feature space representation through metric function, ignoring the information interaction between samples. For complex scene images, similarity and discriminative knowledge need to be summarized from the multiple positive and negative pairs. We propose a novel efficient multi-sample contrastive network (EMSCNet) to integrate knowledge from multiple samples. Specifically, we construct a dynamic dictionary with momentum updates to mine positive and negative pairs from the entire dataset. Then, the similarity and discriminative knowledge between samples are summarized by introducing a contrastive module. Finally, the knowledge of the contrastive module is transferred to the backbone classifier through knowledge distillation. The proposed contrastive module can be easily embedded into the training process of CNNs or ViT and removed during inference. Experimental results conducted on three datasets demonstrate the effectiveness of the proposed approach.

Abstract The traditional visual multi‐object tracking methods based on the Gaussian mixture probability hypothesis density filter are generally not well adapted for tracking the targets in the complex scenarios, where there are a large number of unknowable newborn objects and occluded objects, even some missing objects cannot be associated with their previous trajectories when they are redetected. An improved visual multi‐object tracking algorithm is proposed by integrating an improved efficient convolution operator of the correlation filter and the Gaussian mixture probability hypothesis density filter. First, a similarity matrix based on the intersection‐of‐union is proposed for classifying the objects of survival objects, newborn objects, and then the improved efficient convolution operator method is employed to further identify whether the objects disappear or are missing. Moreover, the feature pyramid similarity is proposed to update the objects for enhancing the tracking accuracy. Finally, compared with some challenging methods on some challenging video sequences from publicly available MOT17 dataset, the proposed Gaussian mixture probability hypothesis density–feature pyramid similarity—efficient convolution operator* method has a good performance on detecting the newborn objects, occluded objects, blurring objects and re‐identifying the missing objects with higher multiple object tracking accuracy.

Hyperspectral image super-resolution addresses the problem of fusing a low-resolution hyperspectral image (LR-HSI) and a high-resolution multispectral image (HR-MSI) to produce a high-resolution hyperspectral image (HR-HSI). Tensor analysis has been proven to be an efficient method for hyperspectral image processing. However, the existing tensor-based methods of hyperspectral image super-resolution like the tensor train and tensor ring decomposition only establish an operation between adjacent two factors and are highly sensitive to the permutation of tensor modes, leading to an inadequate and inflexible representation. In this paper, we propose a novel method for hyperspectral image super-resolution by utilizing the specific properties of high-order tensors in fully-connected tensor network decomposition. The proposed method first tensorizes the target HR-HSI into a high-order tensor that has multiscale spatial structures. Then, a coupled fully-connected tensor network decomposition model is proposed to fuse the corresponding high-order tensors of LR-HSI and HR-MSI. Moreover, a weighted-graph regularization is imposed on the spectral core tensors to preserve spectral information. In the proposed model, the superiorities of the fully-connected tensor network decomposition lie in the outstanding capability for characterizing adequately the intrinsic correlations between any two modes of tensors and the essential invariance for transposition. Experimental results on three data sets show the effectiveness of the proposed approach as compared to other hyperspectral image super-resolution methods.

Hyperspectral image (HSI) fusion refers to the reconstruction of a high-resolution HSI by fusing a low-resolution HSI (LR-HSI) and a high-resolution multispectral image (HR-MSI) over the same scene. Recently, researchers have proposed many approaches to handle this issue. However, most of them assume that both the spatial and spectral degradation functions are known, which are often limited or unavailable in reality. This article presents a novel model-driven deep network based on matrix decomposition, which considers spectral correlations and reasonably embeds the well-known observation models. Specifically, the proposed method decomposes the desired HSI into spectral basis and coefficients. The spectral basis can be estimated from the LR-HSI via singular value decomposition. To learn the coefficients, a learning model is constructed by merging the observation models, matrix decomposition, and sparsity into a concise single formulation. For solving the proposed model, a deep framework is built by unrolling the alternating direction method of multipliers (ADMM), dubbed as ADMM-HFNet, where the involved parameters can be learned adaptively. It is worth noting that the spectral basis cannot fully represent the desired HSI. Therefore, another model is constructed here to supplement the approximation error, which can also be embedded in the deep network. After checking on three datasets, it is found that the proposed method stands out from advanced competing techniques in both quality measures and visual effects.

It is an open question to learn an over-complete dictionary from a limited number of face samples, and the inherent attributes of the samples are underutilized. Besides, the recognition performance may be adversely affected by the noise (and outliers), and the strict binary label based linear classifier is not appropriate for face recognition. To solve above problems, we propose a virtual samples based robust block-diagonal dictionary learning for face recognition. In the proposed model, the original samples and virtual samples are combined to solve the small sample size problem, and both the structure constraint and the low rank constraint are exploited to preserve the intrinsic attributes of the samples. In addition, the fidelity term can effectively reduce negative effects of noise (and outliers), and the ε-dragging is utilized to promote the performance of the linear classifier. Finally, extensive experiments are conducted in comparison with many state-of-the-art methods on benchmark face datasets, and experimental results demonstrate the efficacy of the proposed method.

Abstract The probability hypothesis density (PHD) filter and its cardinalised version PHD (CPHD) have been demonstratedasa class of promising algorithms for multi‐target tracking (MTT) with unknown,time‐varying number of targets. However, these methods can only be used in MTT systems with some prior information of multipletargets, such asdynamic model, newborn target distribution etc.;otherwise, the tracking performance will decline greatly. To solve this problem,an adaptive Grid‐driven technique is proposed based on the framework of the PHD/CPHD filter to recursively estimate the target states without knowing the dynamic model and the newborn target distribution. The grid size can be adaptively adjusted according to the grid resolution, and the dynamic tendencies of the grids can respond to the unknown dynamic models of each targets, including arbitrary manoeuvring models. The newborn targets outside the grid area can be identified by analysing the measurements, and some new grids are generated around them. The experimental results show that the proposed algorithm has a better performance than the traditional particle filter‐based PHD method in terms of average optimal sub‐pattern assignment distance and average target number estimation for tracking multiple targets with unknown dynamic parameters and unknown newborn target distribution.

The dual problem of image super‐resolution (SR), which is referred to as compact‐resolution (CR), and the corresponding image reconstruction are studied. These two problems have been studied independently by the researchers. In this study, a novel model for image CR and the corresponding reconstruction using the reversible network has been proposed. The reversible network has two properties, the first property, lossless information forwarding, which makes the compact‐resolved image retain more information from the original HR image. The second property, bidirectional mapping, by which the forward and reverse propagation of a reversible network can be utilised to implement image CR and reconstruction, respectively, i.e. using the reverse process of image CR to guide the reconstruction. In addition, the utilisation of a reversible network may reduce the size of the model. The superiority of the proposed model was demonstrated by comparing its performance with the state‐of‐the‐art methods on four well‐known benchmark datasets.

In this paper, we investigate the ridge regression for multivariate labels by modelling each pixel and its surrounding pixels as a 3D tensor, and thereby propose a tensor ridge regression approach (TRR) for spatial-spectral hyperspectral image classification. Compared with the traditional ridge regression model, not only the spatial information is incorporated, but also the intrinsic spatial-spectral structure is captured. Moreover, the proposed TRR method is universal that it can be adopted to deal with the fusion of multiscale features for classification purpose. Experiment results conducted on two hyperspectral scenes demonstrate the effectiveness of the proposed method.

We propose an adaptive visual target tracking algorithm based on Label-Consistent K -Singular Value Decomposition (LC-KSVD) dictionary learning. To construct target templates, local patch features are sampled from foreground and background of the target. LC-KSVD then is applied to these local patches to simultaneously estimate a set of low-dimension dictionary and classification parameters (CP). To track the target over time, a kernel particle filter (KPF) is proposed that integrates both local and global motion information of the target. An adaptive template updating scheme is also developed to improve the robustness of the tracker. Experimental results demonstrate superior performance of the proposed algorithm over state-of-art visual target tracking algorithms in scenarios that include occlusion, background clutter, illumination change, target rotation and scale changes.

In this paper, a high spatial resolution (HR) hyperspectral image is inferred from a low spatial resolution (LR) hyperspectral image and a HR panchromatic image by taking advantage of the sparse representation pansharpening (SRP) method. Different from the conventional SRP or joint SRP (JSRP) method, this paper proposes a multitask JSRP method for hyperspectral pansharpening, in order to improve the generalization performance of the model. First, multiple HR/LR dictionary pairs are generated by partitioning the multiple features of the panchromatic image and their corresponding downsampled LR versions into patches. Second, the patch-level sparse representation coefficients of the multiple LR hyperspectral image features are jointly estimated under the multiple LR dictionaries. Finally, the estimated sparse representation coefficients are utilized to reconstruct the HR patches under the original HR dictionary, and the desired HR hyperspectral image is obtained by aggregating the HR patches. Experimental results conducted on two hyperspectral scenes demonstrate the effectiveness of the proposed method.

For the problem that the limited star map field angle cannot obtain the complete star map accurately, the paper study astral intrinsic and imaging features, a star map stitching algorithm based on the principle of visual perception is proposed firstly. The matching models of time and space dimensions is constructed by simulating the visual perception, then the stars and the planets points are saved by searching the matching star group dynamically, the star map is stitched and reconstructed efficiently by creating the computer sparse storage model. The experimental results show that the algorithm can achieve data compression quickly, compression ratio is 99.54%, which can reduce complexity of manual processing and can achieve star map stitching accurately.

This paper presents a new spatial-spectral classification method for hyperspectral images, which consists of three main techniques. Firstly, fully constrained least squares (FCLS) that is common in hyperspectral unmixing is investigated for hyperspectral image classification in kernel Hilbert space. Secondly, the spatial-spectral information of hyperspectral images is exploited to improve the classification performance of kernel-based FCLS (KFCLS) by taking advantage of a weighted H1 norm-based regularization term. Finally, the spatial and label information of training pixels is furthermore incorporated into KFCLS to deal with the scenarios dominated by limited training pixels. Experimental results on two real hyperspectral images demonstrate the effectiveness of the proposed method.

The Gaussian inverse Wishart probability hypothesis density (GIW‐PHD) filter is a promising approach for tracking an unknown number of extended targets. However, it does not achieve satisfactory performance if targets in different sizes are spatially close and manoeuvring because the partitioning methods are sensitive to manoeuvres. To solve this problem, the authors propose the shape selection partitioning (SSP) measurement partitioning algorithm. The proposed algorithm first calculates potential centres and shapes of targets. It then combines each centre with different shapes to divide measurements into subcells. Accordingly, some candidate partitions can be obtained. Finally, it selects the most likely candidate partition and outputs the corresponding subcells. Simulation results show that the application of SSP to the GIW‐PHD filter can achieve better performance when targets are spatially close and manoeuvring, which leads to a lower optimal subpattern assignment distance and a higher accuracy of the sum of weights.