Unsupervised temporal action segmentation is a critical yet challenging task in high-level video understanding, which aims to partition continuous, untrimmed videos into distinct action segments without relying on manual annotations during the training phase. Existing methods typically exploit pre-extracted features for clustering or boundary detection but often suffer from limited accuracy and scalability. In this paper, we reframe unsupervised temporal action segmentation as a video alignment learning task and propose a novel framework with a Frame-Segment Contrastive (FSC) loss. The FSC loss simultaneously encourages fine-grained frame-level alignment and global segment-level coherence, effectively capturing temporal structures in a fully label-free manner. To map the learned representations to semantic categories, we introduce an inference-time exemplar-based label transfer strategy. This approach utilizes the alignment path from a single labeled reference video to generate predictions for target videos. Extensive experiments conducted on five benchmark datasets—Pouring, Penn Action, Breakfast, IKEA Assembly, and Desktop Assembly—demonstrate that our method outperforms state-of-the-art approaches, validating its robustness and scalability in modeling complex temporal dependencies.

Using images captured by cameras with different light spectrum sensitivities, training a unified model for cross-spectral scene representation is challenging. Recent advances have shown the possibility of jointly optimizing cross-spectral relative poses and neural radiance fields using normalized cross-device coordinates. However, such method suffers from cross-spectral misalignment when collecting data asynchronously from devices and lacks the capability to render in real-time or handle large scenes. We address these issues by proposing cross-spectral Gaussian Splatting with spatial occupancy consistency, strictly aligns cross-spectral scene representation by sharing explicit Gaussian surfaces across spectra and separately optimizing each view's extrinsic using a matching-optimizing pose estimation method. Additionally, to address field-of-view differences in cross-spectral cameras, we improve the adaptive densify controller to fill non-overlapping areas. Comprehensive experiments demonstrate that SOC-GS achieves superior performance in novel view synthesis and real-time cross-spectral rendering.

Abstract Images can be corrupted during capture or transmission due to clouds, overlaps, and other interferences, deviating from their original state. Image inpainting techniques restore such images, but different types—Synthetic Aperture Radar (SAR), RGB, and infrared—require varying field‐of‐view sizes. SAR and infrared images, with less information, need a larger field of view, leading to uncorrelated interference in distant areas. RGB images, richer in information, are constrained by a limited local field of view, hindering access to full semantic details. To address these challenges, an aggregated convolution progressive network is proposed. This model employs a coarse‐grained inpainting module for initial restoration, enhanced by an aggregated convolution module to capture contextual information. Local and global details are then used to refine the output, improving restoration quality. Additionally, existing datasets predominantly focus on RGB images, lacking diversity. To bridge this gap, a comprehensive dataset covering SAR, RGB, and infrared images under cloud, overlap, and corruption conditions is constructed. This method achieves superior performance, with MAE of 0.05, SSIM of 0.95, and PSNR of 36.68 within a 20–30% mask size range, outperforming state‐of‐the‐art techniques across diverse image types and size ranges. Experimental results validate its effectiveness in advancing image inpainting.

In recent years, with the development of artificial neural networks, efficiently training models for synthetic aperture radar (SAR) image classification tasks has garnered significant attention from researchers. Particularly when dealing with datasets containing a large number of redundant samples, the selection of training samples becomes crucial for efficient model training. To address this, this paper proposes a SAR image quality evaluation‐based training sample selection method, which integrates multiple indicators. First, a comprehensive SAR image quality evaluation index system is established, and then a SAR image quality evaluation model is constructed by combining representative quality evaluation metrics to guide sample selection. Experimental results demonstrate that the proposed method exhibits strong generalization capabilities on two datasets, MSTAR and OpenSarShip, effectively selecting efficient training samples.

Single object tracking (SOT) in complex scenes presents significant challenges in computer vision . In recent years, transformer has shown its demonstrated efficacy in visual object tracking tasks, due to its capacity to capture the long-range dependencies between image pixels. However, two limitations hinder the performance improvement of transformer-based trackers. Firstly, transformer splits and partitions the image into a sequence of patches, which disrupts the internal structural information of the object. Secondly, transformer-based trackers encode the target template and search region together, potentially leading to confusion between the target and background during feature interaction. To address the above issues, we propose a fully transformer-based tracking framework via learning structural prior information, called SPformer. In other words, a self-attention spatial-prior generative network is established for simulating the spatial associations between features. Moreover, the cross-attention structural prior extractors based on Gaussian and arbitrary distributions are developed to seek the semantic interaction features between the object template and the search region, effectively mitigating feature confusion. Extensive experiments on eight prevailing benchmarks demonstrate that SPformer outperforms existing state-of-art (SOAT) trackers. We further analyze the effectiveness of the two proposed prior modules and validate their application in target tracking models.

Remote sensing hyperspectral images are very useful for land cover classification because of their rich spatial and spectral information. However, hyperspectral image acquisition and pixel labeling are laborious and time-consuming, so few-shot learning methods are considered to solve this problem. Deep learning has gradually been used for few-shot hyperspectral classification, but there are some problems. The feature extraction network based on deep learning requires too many parameters to be trained, resulting in a huge network model, which is not conducive to deployment on remote sensing data acquisition equipment. Moreover, due to the lack of label samples, the algorithm based on deep learning is more prone to overfitting. To solve the above problems, considering the advanced characteristics of the kernel method in dealing with nonlinear, small sample and high-dimensional data, we propose a small scale high precision network called 3DCRFF based on the random Fourier feature (RFF) kernel approximation, which is the 3D convolution random Fourier feature network. Firstly, we combine 3D convolution with random Fourier features as the basic structure of the network to extract the spatial and spectral features of HSI cubes. Secondly, we use a classifier based on attention mechanism to classify feature vectors to obtain recognition probability. Finally, the network parameters are solved from the perspective of Bayesian optimization, and the synthetic gradient optimization method is designed and implemented to realize the fast learning of the network. A large number of experiments HSI classification experiments were performed on UP, PC, IP, and Salinas standard remote sensing data sets, the results show that our algorithm outperforms most state-of-the-art algorithms on few-shot classification.

Recent advances in deep convolution neural networks (CNNs) boost the development of video salient object detection (SOD), and many remarkable deep-CNNs video SOD models have been proposed. However, many existing deep-CNNs video SOD models still suffer from coarse boundaries of the salient object, which may be attributed to the loss of high-frequency information. The traditional graph-based video SOD models can preserve object boundaries well by conducting superpixels/supervoxels segmentation in advance, but they perform weaker in highlighting the whole object than the latest deep-CNNs models, limited by heuristic graph clustering algorithms. To tackle this problem, we find a new way to address this issue under the framework of graph convolution networks (GCNs), taking advantage of graph model and deep neural network. Specifically, a superpixel-level spatiotemporal graph is first constructed among multiple frame-pairs by exploiting the motion cues implied in the frame-pairs. Then the graph data is imported into the devised multi-stream attention-aware GCN, where a novel Edge-Gated graph convolution (GC) operation is proposed to boost the saliency information aggregation on the graph data. A novel attention module is designed to encode the spatiotemporal sematic information via adaptive selection of graph nodes and fusion of the static-specific and the motion-specific graph embedding. Finally, a smoothness-aware regularization term is proposed to enhance the uniformity of salient object. Graph nodes (superpixels) inherently belonging to the same class will be ideally clustered together in the learned embedding space. Extensive experiments have been conducted on three widely used datasets. Compared with fourteen state-of-the-art video SOD models, our proposed method can well retain the salient object boundaries and possess a strong learning ability, which shows that this work is a good practice for designing GCNs for video SOD.

Video salient object detection aims at distinguishing the salient objects from the complex background and highlighting them uniformly in the spatiotemporal domain, which still suffers from the interference of the complicated dynamic background in unconstrained videos. To address this problem, we propose a novel coarse-to-fine spatiotemporal salient object detection method. Specifically, we first model a novel motion energy to exclude the motion noise by exploiting the motion magnitude and motion orientation. Then, a supervoxel-level inter-frame graph model is constructed for each pair of adjacent frames independently, and a robust graph clustering-based saliency seed generation method is proposed to produce a coarse saliency map. Furthermore, the supervoxel-level inter-frame graph model is reconstructed by considering the regional spatiotemporal consistency constraint based on the coarse saliency map. The prior information obtained from pixel clustering is also taken into account to optimize the weight of the inter-frame graph model. Finally, a multi-graphs saliency propagation method is exploited under the manifold regularization framework by fusing the motion energy and appearance feature to refine the coarse saliency map. The extensive experiments on two widely used datasets validate the effectiveness and superiority of the proposed method against 13 state-of-the-art methods in terms of PR-curves, scores of S-measure,$F_{\beta }$, and MAE.

Computer vision based detection approaches are widely employed to detect or calibrate different types of meters nowadays. However, traditional detection algorithms suffer drawbacks in accuracy and adaptability upon detecting various types of automobile dashboards. Plenty of parameters of these algorithms need to be tuned to suit certain types of dashboards. Besides, theses algorithms cannot automatically read the speed value, which requires manual setting operations. In this paper, a novel approach is presented to adaptively detect different types of automobile dashboards. The contour analysis based method is first implemented to extract the connected component of the pointer. A robust character segmentation classifier, which is designed by cascading histogram of oriented gradients (HOG)/support vector machine (SVM) binary classifier, character filter as well as HOG/multiclass SVM digit classifier, is then proposed to recognize digit characters on the dashboard. Simultaneously, tick marks are then extracted based on recognition results. Finally, Newton interpolation linear relationship is established to diagnose the potential responding errors of the pointer. The experimental results show that the pointer extraction method is robust to interferences caused by connected components of digits and also that the established character segmentation classifier has a more accurate detection result. Furthermore, compared with similar algorithms, it has a significant advantage in detecting a vast majority of different dashboards without manual tuning of the parameters.

We present a novel, robust estimation method to distinguish salient objects from complicated, dynamic backgrounds in videos. In this method, we propose a novel approach to model motion energy based on motion magnitude, motion orientation, gradient flow field, and spatial gradient of the video frame. Furthermore, an effective spatiotemporal objectness map is also proposed to estimate a compact object-like region in the current video frame leveraging both the objectness proposals and the saliency map of the previous frame. Then the current video frame is oversegmented into the granularity of superpixels using the simple linear iterative clustering algorithm. Each superpixel is designated as a node of a graph. The similarity between adjacent superpixels will be assigned as the weight of an edge that connects these two nodes. The feature values of motion energy and spatiotemporal objectness within each superpixel will be averaged respectively, and used to graphically cluster similar superpixels to form the detected salient object. Extensive experiments comparing this proposed new method against twelve existing salient object detection (SOD) methods have been performed using the benchmark datasets unconstrained video saliency detection and densely annotated video segmentation. Superior performance of this proposed SOD method has been observed through three well-known performance metrics: precision-recall curves, F-measure curves, and the mean absolute error.

Feature selection in data-driven modelling is an important research topic for prognostics. The performance of prediction model may vary considerably under different feature subsets. Hence it is important to devise a systematic feature selection method, which offers the guidance for choosing the mos t representative features for prognostics. Nowadays, feature selection algorithms in the field of prognostics are largely studied to the type of learning: supervised or unsupervised, which leads to poor generalization between different prognostics applications. In this paper, a unified feature selection method, called locality structure preserving based feature selection (LSPFS), is developed to improve the robustness and accuracy of prognostics under both unsupervised and supervised learning conditions. In LSPFS, the local structure of original data is constructed according to the similarity between data points, and the representative features are selected based on their ability to preserve the local structure. Moreover, by designing different local structure via local information and actual degradation information of the data, the introduced method can unify supervised and unsupervised feature selection, and enable their joint study under a general framework. Experiments on NASA turbofan engine simulation dataset and lithium-ion battery dataset are conducted to test and evaluate the proposed algorithm.

A novel feature extraction method, namely Laplacian discriminant projection with optimized kernels (KLDP-Opt) algorithm is proposed in this paper. The advantage of KLDP-Opt lies in: 1) the similarity matrix is constructed with the class-wise nonparametric similarity measure where it solves procedure selection problem; 2) data-dependent kernel is applied to solve the limitation of linearity of LPP, where the adaptive parameters of the data-dependent kernel are computed through optimizing an objective function designed for measuring the class separability of data in the feature space. Besides the theory derivation, the experiments are implemented on ORL and Yale face databases to evaluate the feasibility of the proposed algorithm.

This paper presents a novel face recognition method based on complete Kernel Fisher discriminant (CKFD) analysis of Gabor features with power polynomial models. By integrating the Gabor wavelet representation of face images and the enhanced powerful discriminator named CKFD analysis, the method is robust to changes in illumination and facial expressions and poses. On the other hand, the extended polynomial Kernels, namely fractional power polynomial (FPP) models, are employed in CKFD analysis, which enhance face recognition performance. Comparing with existing PCA, LDA, KPCA, KFD and CKFD methods, the proposed method gives superior results in the ORL and Yale face databases. Its good performance in the two face databases gives the promising idea to solve the pose, illumination, and expression (PIE) problem of face recognition