Multi-scale super-pixel hyperspectral remote sensing image classification method for coupling spatial-spectral characteristics

Abstract

The invention provides a multi-scale super-pixel hyperspectral remote sensing image classification method for coupling spatial-spectral characteristics. The method comprises the following steps: dividing a hyperspectral remote sensing image data set into a training set and a test set, and performing dimension reduction processing on the training set by using a principal component analysis (PCA) to obtain an effective spectral band; under different scales, carrying out super-pixel segmentation processing on effective spectral bands by using a super-pixel segmentation algorithm ERS based on an entropy rate; calculating the similarity between any two superpixels through an RBF kernel function, and further obtaining a spatial spectrum kernel matrix Kpp of the training set; calculating the similarity between any two pixels in the training set through a polynomial kernel function, and then obtaining an original spectrum kernel matrix Kyp of the training set; fusing the spatial spectrum kernel matrix Kpp and the original spectrum kernel matrix Kyp to obtain a multi-scale super-pixel spatial spectrum synthesis kernel matrix, and then training an SVM classifier model; and classifying the test set by using the trained SVM classifier model, and outputting to obtain a corresponding ground feature classification image.

Description

technical field [0001] The invention relates to the technical field of remote sensing image processing, in particular to a multi-scale superpixel hyperspectral remote sensing image classification method coupled with spatial spectral features. Background technique [0002] Due to its high timeliness, remote sensing images have been used as the main data source for land use classification, medical image processing, target detection and recognition, etc. It improves the distinction and recognition of different species, and is favored by researchers. Hyperspectral remote sensing images contain not only spatial information of ground objects, but also rich spectral information of ground objects. The exploration of hyperspectral image classification methods is of great significance for distinguishing ground objects and grasping regional ground object information in real time. The existing hyperspectral remote sensing image classification research uses a single-scale superpixel met...

AI Technical Summary

Problems solved by technology

[0005] In view of the low classification accuracy of existing hyperspectral remote sensing image classification methods, the present invention proposes a multi-scale super-pixel hyperspectral remote sensing image classification method coupled with spatial spectral features, which can effectively improve the classification accuracy of hyperspectral remote sensing images

Images