Band selection (BS) is a vital technique for improving efficiency of hyperspectral image (HSI) processing. This letter proposes a dual-view structural similarity subspace clustering model (DVS3C) for BS. Traditional low-rank subspace clustering (LRSC) methods rely solely on single-view data (e.g., original HSI), potentially leading to the loss of critical information (e.g., spatial structures) and insufficient exploitation of the multi-dimensional features of HSI for optimal BS. To do so, DVS3C constructs a spatial view alongside the spectral view, leveraging global spectral-spatial information through subspace clustering to achieve complementary advantages between views. Besides, to overcome LRSC’s limitations in capturing band local structure, DVS3C introduces a structural similarity matrix to deeply exploit intraview neighborhood relationships of bands, further reducing band redundancy. Ultimately, an adaptive dual-view fusion strategy that iteratively optimizes a consensus matrix while dynamically adjusting the contribution of each view is designed to ensure view consistency. Experimental results on four public datasets demonstrate its remarkable stability and superiority. The source code is available athttps://github.com/ydk0912/DVS3C.

Contrastive learning has recently demonstrated great potential in hyperspectral image few-shot classification. However, conventional methods mainly emphasize instance-level similarity while neglecting class structure, often leading to the separation of intra-class samples. Moreover, the lack of explicit class prototype modeling often leads to ambiguous decision boundaries. To address these issues, this paper proposes a prototype-guided supervised momentum contrastive learning (PGSMC) for hyperspectral cross-domain few-shot classification. PGSMC first designs an asymmetric augmentation module to enhance sample diversity by applying distinct augmentation strategies and encoders to the query and key views. Subsequently, a momentum queue is employed to store historical key features and their corresponding labels. This mechanism enables smooth updates of class-level momentum prototypes and mitigates the limitations imposed by mini-batch training. Finally, a momentum prototype contrastive loss is formulated to guide the model toward class-level feature representations, thereby promoting more discriminative decision boundaries. Overall, PGSMC enables query samples to contrast with more representative momentum prototypes, enhancing inter-class separability and promoting well-defined class boundaries. Extensive experiments on five hyperspectral image datasets demonstrate that PGSMC significantly outperforms existing few-shot learning methods.

Abstract As one of the dimensionality reduction techniques of hyperspectral image (HSI), band selection (BS) does not change the spectral characteristics and physical meaning of HSIs, which is beneficial to the identification and analysis of surface objects. Recently, many BS methods for target detection have achieved promising results by making full use of the priori spectral features of the target to be detected. Conversely, anomaly detection separates the anomaly based solely on the statistical distribution difference between anomaly and background without any prior information. Therefore, the development of BS for anomaly detection has lagged far behind that of BS for target detection. To this end, this paper proposes a novel BS algorithm dedicated to anomaly detection tasks, named anomaly‐background separation and particle swarm optimization (PSO)‐based BS. Specifically, an anomaly‐background separation framework (ABSF) is established to predetermine a priori knowledge of anomaly distribution. Then, three band prioritization criteria are constructed with the anomaly‐background constraints generated by ABSF. Finally, PSO is used to find the optimal subset of bands in the solution space. The experiments on two real datasets demonstrate that the proposed method yields better detection results and greater stability compared to other BS methods discussed in this paper.

Due to the redundancy and sparsity of hyperspectral data, sparse representation (SR) has proven to be well-suited for hyperspectral band selection (BS). Moreover, graph regularizers can effectively incorporate local structural information of the data to improve the solution of SR. However, existing unsupervised BS approaches typically consider only a simple graph based on a single spectral metric. In contrast, the hypergraph can capture the multiple adjacencies of the bands and has significant advantages. This letter proposes a hypergraph-regularized self-representation model (HyGSR) for BS. HyGSR is innovative in that it jointly combines the spectral similarity and band index as a new similarity metric to rationalize the local structure of bands extracted by hypergraph, while using a robustl2,1-norm to exploit the sparse properties of the data for BS. Experimental results on four real hyperspectral scenarios verify that HyGSR outperforms other competitors with high stability and usability.

Band selection (BS) methods based on deep learning have achieved significant development. However, most existing band selection methods commonly utilize a fully connected neural network (FCN) or convolutional neural network (CNN) to explore the correlation among bands and rarely combine the two styles of the network to select bands. Moreover, almost all the methods employ the form of the combination of$L_{1}$norm and Sigmoid to constitute attention model, which may lead to losing some informative band feature. To tackle these troubles, this letter proposes a novel band selection network using FCN and CNN, termed as global-local spectral weight network based on attention (GLSWA), in which the band features of each pixel is mined using the network of two types, and designing an attention-based scoring module (ASM) and a convolutional reconstruction module (CRM), respectively, so that each attention of band is adjusted by simultaneous considering the entire band features and successive one. Experimental results on three real hyperspectral image (HSI) datasets show that the proposed method achieves satisfactory accuracy than some state-of-the-art algorithms.

Band selection (BS) is considered as an effective method for dimensionality reduction of hyperspectral data. As an important application for hyperspectral remote sensing, target detection is widely concerned. Therefore, how to select more representational band subset for specific target to improve performance of detection is worth discussing. This letter proposed a BS method for specific target detection, called constrained target band selection under adaptive subspace partitioning (CTASPBS). Firstly, all bands are partitioned into multiple weakly correlated subsets via adaptive subspace partition strategy (ASPS). Then, according to a target-constrained band prioritization (BP) criterion, the band with the highest priority in each subset is selected to form the optimal band subset. Due to the application of different BP criterion, two BS methods ASPS_MinV and ASPS_MaxV are proposed. Finally, experimental results on real hyperspectral data show that CTASPBS is an effective BS method for specific target detection.