With the development of deep learning, the use of convolutional neural networks (CNN) to improve the land cover classification accuracy of hyperspectral remote sensing images (HSRSI) has become a research hotspot. In HSRSI semantics segmentation, the traditional dataset partition method may cause information leakage, which poses challenges for a fair comparison between models. The performance of the model based on “convolutional-pooling-fully connected” structure is limited by small sample sizes and high dimensions of HSRSI. Moreover, most current studies did not involve how to choose the number of principal components with the application of the principal component analysis (PCA) to reduce dimensionality. To overcome the above challenges, firstly, the non-overlapping sliding window strategy combined with the judgment mechanism is introduced, used to split the hyperspectral dataset. Then, a PSE-UNet model for HSRSI semantic segmentation is designed by combining PCA, the attention mechanism, and UNet, and the factors affecting the performance of PSE-UNet are analyzed. Finally, the cumulative variance contribution rate (CVCR) is introduced as a dimensionality reduction metric of PCA to study the Hughes phenomenon. The experimental results with the Salinas dataset show that the PSE-UNet is superior to other semantic segmentation algorithms and the results can provide a reference for HSRSI semantic segmentation.

Hyperspectral image (HSI) classification is an important but challenging topic in the field of remote sensing and earth observation. By coupling the advantages of convolutional neural network (CNN) and Transformer model, the CNN–Transformer hybrid model can extract local and global features simultaneously and has achieved outstanding performance in HSI classification. However, most of the existing CNN–Transformer hybrid models use artificially specified hybrid strategies, which have poor generalization ability and are difficult to meet the requirements of recognizing fine-grained objects in HSI of complex scenes. To overcome this problem, we proposed a convolution–Transformer adaptive fusion network (CTAFNet) for pixel-wise HSI classification. A local–global fusion feature extraction unit, called the convolution–Transformer adaptive fusion kernel, was designed and integrated into the CTAFNet. The kernel captures the local high-frequency features using a convolution module and extracts the global and sequential low-frequency information using a Transformer module. We developed an adaptive feature fusion strategy to fuse the local high-frequency and global low-frequency features to obtain a robust and discriminative representation of the HSI data. An encoder–decoder structure was adopted in the CTAFNet to improve the flow of fused local–global information between different stages, thus ensuring the generalization ability of the model. Experimental results conducted on three large-scale and challenging HSI datasets demonstrate that the proposed network is superior to nine state-of-the-art approaches. We highlighted the effectiveness of adaptive CNN–Transformer hybrid strategy in HSI classification.