Hyperspectral image joint classification method based on multi-feature learning and superpixel kernel sparse representation

Abstract

The invention discloses a hyperspectral image joint classification method based on multi-feature learning and superpixel kernel sparse representation. used for improving the classification precision of hyperspectral images. According to the method, spatial features and local and global super-pixel features are extracted firstly, and spatial information is fully utilized to improve the classification precision; then, a kernel sparse representation classification method based on multiple features is provided, and the classification problem that a high-dimensional feature space is linearly inseparable is effectively solved; moreover, the similarity and diversity among multiple features are fully utilized, a multi-kernel learning method is used for constructing a basic kernel and optimizing the coefficient of the basic kernel, and a composite kernel is designed to replace a single kernel in a kernel sparse representation method, so that the classification precision is further improved. Compared with similar classic hyperspectral image classification methods, the classification method is higher in classification precision and can be applied to the fields of hyperspectral classification,aerospace remote sensing and the like.

Description

technical field [0001] The invention relates to a hyperspectral image joint classification method based on multi-feature learning and superpixel kernel sparse representation, which belongs to the technical field of hyperspectral data processing, is used to effectively improve the classification accuracy in the case of small samples, and can be applied to aerospace remote sensing detection and other practical applications engineering field. Background technique [0002] Hyperspectral image classification technology is a hot research issue in the field of remote sensing, and its goal is to classify each spectral pixel into a specific category. In recent years, many classification methods have been proposed, including MLR, SVM and neural network. However, these methods only exploit the spectral information of hyperspectral data and ignore the spatial information, thus leading to noisy classification results. In order to overcome this problem, many methods that try to combine ...

AI Technical Summary

Problems solved by technology

Existing multi-feature-based classification methods still have obvious problems, such as how to fully consider the similarity and diversity among different features

Images