To improve the accuracy and generalization ability of hyperspectral image classification, a feature extraction method integrating principal component analysis (PCA) and local binary pattern (LBP) is developed for hyperspectral images in this article. The PCA is employed to reduce the dimension of the spectral features of hyperspectral images. The LBP with low computational complexity is used to extract the local spatial texture features of hyperspectral images to construct multifeature vectors. Then, the gray wolf optimization algorithm with global search capability is employed to optimize the parameters of kernel extreme learning machine (KELM) to construct an optimized KELM model, which is used to effectively realize a hyperspectral image classification (PLG-KELM) method. Finally, the Indian pines dataset, Houston dataset, and Pavia University dataset and an application of WHU-Hi-LongKou dataset are selected to verify the effectiveness of the PLG-KELM. The comparison experiment results show that the PLG-KELM can obtain higher classification accuracy, and takes on better generalization ability for small samples. It provides a new idea for processing hyperspectral images.

Recently, the rapid development of deep learning has greatly improved the performance of image classification. However, a central problem in hyperspectral image (HSI) classification is spectral uncertainty, where spectral features alone cannot accurately and robustly identify a pixel point in a hyperspectral image. This paper presents a novel HSI classification network called MS-RPNet, i.e., multiscale superpixelwise RPNet, which combines superpixel-based S³-PCA with two-dimensional singular spectrum analysis (2D-SSA) based on the Random Patches Network (RPNet). The proposed frame can not only take advantage of the data-driven method, but can also apply S³-PCA to efficiently consider more global and local spectral knowledge at the super-pixel level. Meanwhile, 2D-SSA is used for noise removal and spatial feature extraction. Then, the final features are obtained by random patch convolution and other steps according to the cascade structure of RPNet. The layered extraction superimposes the different sparial information into multi-scale spatial features, which complements the features of various land covers. Finally, the final fusion features are classified by SVM to obtain the final classification results. The experimental results in several HSI datasets demonstrate the effectiveness and efficiency of MS-RPNet, which outperforms several current state-of-the-art methods.

Hyperspectral image (HSI) in Internet of Things (IoT) is a typical small sample dataset, which is difficult and costly to label samples manually. In feature extraction, it is difficult to increase the inter-class distance and reduce the intra-class variance according to the limited label information, resulting in easy misclassification of the extracted features. To solve this problem, this paper proposes an adaptive pseudo-label feature learning (APFL) model. In the APFL model, a hybrid distance pseudo-label generation (HDPG) method was designed to generate pseudo-labels by iterative multi-scale superpixel segmentation using spectral-spatial mixing distance information, while a pseudo-label feature generation (PFG) method was designed to generate pseudo-label features using pseudo-labels to capture the intra-class average vectors of HSI principal component features. Finally, the extracted pseudo-label features are classified at pixel level. This APFL model can effectively reduce the intra-class variance and increase the inter-class distance of HSI data, thus improving the inter-class separability. We have done comparative verification experiments on five commonly used HSI datasets in IoT. Compared with the current advanced feature extraction methods and classification methods, the proposed APFL model in this paper has higher classification accuracy.

In the field of hyperspectral image classification, using spatial information as a supplement to spectral information has been widely applied. This paper proposes a novel dual-scale complementary spatial-spectral joint classification model (DSCSM) to mitigate the issues of detail loss and insufficient utilization of spatial information, which traditionally lead to lower classification accuracy. In essence, the final classification result is obtained through decision fusion of two complementary feature extraction stages. In the pre-processing stage, a new dual-scale truncated filtering feature extraction method (DTFE) is proposed, which uses truncated filters with two different parameter settings to obtain two scales of smoothed patches, and then fuses them to obtain dual-scale structural features using Kernel Principal Component Analysis (KPCA). DTFE preserves edge information while smoothing details, effectively removing noise and retaining the dual-scale feature information. In the post-processing stage, a sub-Markov random walk (subRW) based spatial probability optimization method (SPO) is proposed, which models the spatial association of neighboring pixels, retaining complex textures as well as weak edge information to optimize the classification probability. Finally, the decision fusion strategy is employed to integrate the classification probabilities acquired from the aforementioned two stages. Comparative experiments on six different scene datasets with state-of-the-art classification methods validate that even with a small number of samples, DSCSM can achieve excellent object recognition performance. In addition, comprehensive parameter analysis proves the robustness and stability of the proposed method.

Hyperspectral image (HSI) spectral-spatial joint feature extraction methods generally suffer from low feature retention and weak spatial-spectral dependence, which will lead to single class feature confrontation (SCFC). To solve this problem, an unsupervised multivariate feature fusion network (M ³ FuNet) is developed in this paper. In M ³ FuNet, multiscale supervector matrix correction (MSMC) and multiscale random convolution dispersion (MRCD) is used as the spectral and spatial feature extraction method, and the feature retention of spectral and spatial features is improved to achieve feature calibration by feature fusion and decision fusion, called multivariate feature fusion. The MSMC is employed to correct supervector matrix to reduce the intraclass variance in superpixel homogeneous regions, overcome the phenomenon of supervector block drift (SvBD). The MRCD uses random convolution and gaussian smoothing to extract deep spatial features. Due to the similar feature representation ability of the MSMC and MRCD, the obtained spectral-spatial joint features have high feature retention and strong spectral-spatial dependence. Finally, this multivariate feature fusion network is used for realizing classification of HSI. Three commonly HSI datasets are used to validate the effectiveness of the M ³ FuNet. The experiment results show that the M ³ FuNet has more superior performance by comparing with several state-of-the-art HSI classification methods. The code of the proposed M ³ FuNet is available at http://github.com/aichou233/M3FuNet.