Hyperspectral image clustering method based on residual subspace clustering network

Abstract

The invention provides a hyperspectral image clustering method based on a residual subspace clustering network, and the method specifically comprises the following steps: carrying out the preprocessing of an original hyperspectral image, and obtaining a standardized spatial spectrum sample; constructing a residual subspace clustering network, inputting the standardized empty spectrum sample into the residual subspace clustering network, and performing parameter optimization by adopting a gradient descent method to obtain a nonlinear self-expression coefficient matrix C; constructing a similarity matrix A by using the nonlinear self-expression coefficient matrix C, and segmenting the similarity matrix A into k groups by using a spectral clustering algorithm so as to obtain k clustered clusters. The method has the beneficial effects that a linear subspace clustering method is expanded to a nonlinear depth model, and the problem of low unsupervised classification precision of the hyperspectral image is effectively solved by learning clustering-oriented depth feature representation.

Description

technical field [0001] The invention relates to the field of hyperspectral image classification, in particular to a hyperspectral image clustering method based on a residual subspace clustering network. Background technique [0002] Hyperspectral image classification is an important basis for hyperspectral remote sensing applications. According to whether labeled samples are used, hyperspectral image classification can be divided into supervised classification and unsupervised classification. In the past ten years, hyperspectral image supervised classification methods have achieved great success, such as support vector machines, random forests, and extreme learning machines. Because supervised classification uses labeled information, it can usually obtain higher accuracy, but it also requires a lot of manpower and material resources to label samples, resulting in high training costs for supervised classification methods, which is not conducive to the application in productio...

AI Technical Summary

Problems solved by technology

Because supervised classification uses labeled information, it can usually obtain higher accuracy, but it also requires a lot of manpower and material resources to label samples, resulting in high training costs for supervised classification methods, which is not conducive to the application in production practice.

[0003] The deep clustering method has not been applied in hyperspectral image clustering, and there are two problems. One is that the global structure information of the sample data is not considered, and the other is that the training is difficult and needs to be combined with pre-training.

Images