An Improved Hyperspectral Image Classification Method

Abstract

The invention discloses an improved hyperspectral image classification method. The method includes the following steps: (1) Spatial neighbor selection of the central pixel of the hyperspectral image to be classified: using a watershed segmentation region selection strategy or a minimum spanning tree neighbor selection strategy to obtain a high-quality spatial neighbor region; (2) selecting a high-quality spatial neighbor region Perform spatial feature extraction; (3) Spectral-space joint model prediction: use the synthetic kernel method to fuse spatial and spectral features, and then train the classification model to predict the hyperspectral image test set label. The present invention adopts different selection methods of spatial neighbors to meet the different requirements for speed and precision in hyperspectral classification; in addition, by mining and utilizing spatial information, it effectively solves the problems of the same object with different spectra and the same spectrum in hyperspectral images. Foreign matter problem, high-quality spatial neighbors and feature information enhance the robustness of the original spectral classification, so it has high use value.

Description

technical field [0001] The invention belongs to the field of hyperspectral image processing, in particular to an improved hyperspectral image classification method. Background technique [0002] With the continuous development of remote sensing hardware, hyperspectral remote sensing image processing technology has been developed rapidly and widely used, attracting the attention of a large number of researchers. Traditional remote sensing image classification only uses less band spectral information, while hyperspectral images contain hundreds of band spectral information, which is more helpful for classification, but classical pattern recognition methods classify them, and the misclassification phenomenon is less Seriously, the effect is not ideal. [0003] The rich spectral information in hyperspectral images also contains many challenges and problems, such as high-dimensional small sample classification, same object with different spectra, same spectrum with different obj...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the technical problems of the same object with different spectra and different objects with the same spectrum in hyperspectral images, the present invention proposes an improved hyperspectral image classification method, which can provide reliable information for real-time response and high-precision application scenarios. Hyperspectral Image Classification

Images