Despite success in the same-scene hyperspectral image classification (HSIC), for the cross-scene classification, samples between source and target scenes are not drawn from the independent and identical distribution, resulting in significant performance degradation. To tackle this issue, a novel unsupervised domain adaptation (UDA) framework towards multi-level features and decision boundaries (ToMF-B) is proposed for the cross-scene HSIC, which can align task-related features and learn task-specific decision boundaries in parallel. Based on the maximum classifier discrepancy, a two-stage alignment scheme is proposed to bridge the interdomain gap and generate discriminative decision boundaries. In addition, to fully learn task-related and domain-confusing features, a CNN and Transformer-based multi-level features extractor (generator) is developed to enrich the feature representation of two domains. Furthermore, to alleviate the harm even the negative transfer to UDA caused by task-irrelevant features, a task-oriented feature decomposition method is leveraged to enhance the task-related features while suppressing task-irrelevant features, enabling the aligned domain-invariant features can be contributed to the classification task explicitly. Extensive experiments on three cross-scene HSI benchmarks have validated the effectiveness of the proposed framework.

Deep learning (DL) based methods represented by convolutional neural networks (CNNs) are widely used in hyperspectral image classification (HSIC). Some of these methods have strong ability to extract local information, but the extraction of long-range features is slightly inefficient, while others are just the opposite. For example, limited by the receptive fields, CNN is difficult to capture the contextual spectral-spatial features from a long-range spectral-spatial relationship. Besides, the success of DL-based methods is greatly attributed to numerous labeled samples, whose acquisition are time-consuming and cost-consuming. To resolve these problems, a hyperspectral classification framework based on multi-attention Transformer (MAT) and adaptive superpixel segmentation-based active learning (MAT-ASSAL) is proposed, which successfully achieves excellent classification performance, especially under the condition of small-size samples. Firstly, a multi-attention Transformer network is built for HSIC. Specifically, the self-attention module of Transformer is applied to model long-range contextual dependency between spectral-spatial embedding. Moreover, in order to capture local features, an outlook-attention module which can efficiently encode fine-level features and contexts into tokens is utilized to improve the correlation between the center spectral-spatial embedding and its surroundings. Secondly, aiming to train a excellent MAT model through limited labeled samples, a novel active learning (AL) based on superpixel segmentation is proposed to select important samples for MAT. Finally, to better integrate local spatial similarity into active learning, an adaptive superpixel (SP) segmentation algorithm, which can save SPs in uninformative regions and preserve edge details in complex regions, is employed to generate better local spatial constraints for AL. Quantitative and qualitative results indicate that the MAT-ASSAL outperforms seven state-of-the-art methods on three HSI datasets.

Convolutional neural networks are widely used in the field of hyperspectral image classification because of their excellent nonlinear feature extraction ability. However, as the sampling position of the regular convolution kernel is unchangeable, the regular convolution cannot distinctively extract the spatial and spectral information around the central pixel, which makes the classification results at the boundaries of ground objects over-smoothed and the classification performance degraded. Thus, we propose a novel superpixel guided deformable convolution network (SGDCN) for hyperspectral image classification. Firstly, the superpixel region fusion filter (SRF-Filter) is designed to fuse the initial superpixel region segmented by the simple linear iterative clustering (SLIC), making the fused superpixel region have a high homogeneity and also contain spatial features of diverse scales. Then, the superpixel guided deformable convolution (SGD-Conv) is proposed to make the shape of deformable convolution consistent with the real shape of land covers, and the SGD-Conv can extract pure neighborhood spatial-spectral features. Finally, a superpixel joint bilateral filter (SPJBF) is designed to solve the pixel-level and region-level misclassification problem, which can effectively utilize the superpixel region’s homogeneity and improve the classification accuracy. Experiments on three HSI datasets indicate that the SGDCN can obtain better classification performance when compared with other twelve state-of-the-art methods.

Nowadays, graph convolutional networks (GCNs) are getting more attention in hyperspectral image classification (HSIC), and various algorithms based on GCNs have been proposed. However, because of hyperspectral images’ (HSIs) complex spatial texture information, the long-range graph convolution (GConv) and short-range GConv may cause inaccurate or oversmoothed feature extraction of some nodes. Thus, a multiscale short- and long range graph convolutional network (MSLGCN) is proposed for HSIC. First, MSLGCN not only extracts spatial information of ground objects at different scales but also simultaneously captures global and local spectral features, which preserves objects’ fine boundaries. Then, the rich multiscale information is complementary, enabling the MSLGCN to take full advantage of texture structures of varying sizes. In addition, a method to determine the superpixel scale by the intrinsic properties of HSIs is proposed to ensure that the segmentation boundary depicts the texture structure of the object accurately. Finally, the short–long graph convolution (SLGConv) is designed to fuse the advantages of global and local features, enabling the MSLGCN to extract accurate spatial–spectral features of nodes at any location. Experiments on three HSI datasets indicate that the MSLGCN can obtain better classification performance when compared with the other 11 state-of-the-art methods.

Nowadays, domain adaptation (DA) is getting more attention in cross-scene hyperspectral image (HSI) classification, and various DA algorithms have been proposed. However, regular convolution indiscriminately extracting features around the center pixel will result in the inaccurate extraction of spatial-spectral features, which significantly affect the subsequent feature alignment. Meanwhile, the method of aligning the category features of source and target domains from a single-level may not cope well with complex HSIs. Therefore, we propose a multilevel feature alignment algorithm based on spatial attention deformable convolution (MFA-SADC), which achieves multilevel feature alignment from feature to feature, feature to cluster-center, and cluster-center to cluster-center. In addition, spatial attention deformable convolution is proposed to compose the feature extraction network of MFA-SADC, which guarantees the purity of spatial-spectral features. Experiments on three HSI datasets indicate MFA-SADC can obtain better classification performance when compared with the seven state-of-the-art methods.

Nowadays, foundation models have demonstrated exceptional performance across numerous downstream tasks. However, the effective application of these models to hyperspectral image classification (HSIC) is challenged by the unique characteristics of hyperspectral data, including high dimensionality, high variability, and high spatial structure complexity. Therefore, methods need to be developed that leverage the advantages of foundation models while addressing these challenges. Firstly, a novel HSIC algorithm based on a frozen-parameter Segment Anything Model (SAM) encoder, called SAGFFNet, is proposed. This framework represents the first attempt to use a frozen SAM encoder for global feature extraction and to use spectral dimension data as prompts, enabling precise global spatial-spectral feature extraction with the aid of spectral information. Secondly, by introducing the self-adaptive padding mechanism and the global feature extraction subnetwork (GFEsNet), the model is enabled to extract distinctive and discriminative features for each category from hyperspectral data through varying padding sizes, thereby enhancing the feature extraction and generalization capabilities of the foundation model. Subsequently, the spectral feature prompt subnetwork (SFPsNet) is designed to extract spectral feature information from samples of different classes as prompt features, assisting the framework in better understanding the global features extracted by GFEsNet. Finally, the semantic information decoder subnetwork (SIDsNet) is introduced as a semantic information decoder, achieving efficient fusion of global spatial-spectral features and spectral prompt features, which significantly improves classification performance. Experiments conducted on four hyperspectral image datasets show that the proposed method outperforms nine existing approaches in terms of classification accuracy.