Hyperspectral image classification method based on a principal component analysis network and spatial coordinates

Abstract

The invention discloses a hyperspectral image classification method based on a principal component analysis network and spatial coordinates, and mainly solves the problems that in the prior art, the fusion of space and spectral information is complex or insufficient, and the computational complexity is high when the principal component analysis network is used for hyperspectral classification. According to the implementation scheme, the method comprises the following steps: reading a data set of a hyperspectral image; randomly selecting a training set and a test set from the data set accordingto spatial partitioning; then carrying out dimension reduction, normalization and edge reservation filtering processing on the spectral information; expanding the spatial coordinates and fusing the spatial coordinates with the spectral characteristics; training a principal component analysis network to obtain a trained principal component analysis network; inputting the test set data into the trained principal component analysis network to obtain a feature vector of each pixel in the test set; and finally, obtaining a classification result by using a support vector machine (SVM). According tothe method, the calculation complexity is reduced, the classification effect is improved, and the method can be applied to target recognition in resource exploration, forest coverage and disaster monitoring.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a hyperspectral image classification method, which can be applied to target recognition in resource exploration, forest coverage, and disaster monitoring. Background technique [0002] The key to hyperspectral image classification technology is to use a small number of training samples to obtain higher classification accuracy. In the early days, hyperspectral images were mainly classified using spectral information. In recent years, researchers have found that spatial information of hyperspectral images is also very important. Therefore, how to simultaneously and fully utilize spectral information and spatial information has become an The key to precision. [0003] In their paper "R-VCANet: A New Deep-Learning-Based Hyperspectral Image Classification Method" (IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2017, 10(5): 1975-1986), P...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that the use of spatial information is not sufficient, that is, only the spatial information of the hyperspectral image is used when filtering; in addition, this method inputs all dimensional spectral information into the improved VCANet network for training , the computational complexity is large

The disadvantage of this method is that the spatial information of the hyperspectral image is underutilized, and only the spatial coordinates are used, and the spatial coordinates are not good enough for the classification of ground object categories where the sample distribution is not concentrated or the sample size is small.

Images