Hyperspectral Image Classification Method Based on Low Rank-Sparse Information Combination Network

Abstract

The invention discloses a hyperspectral image classification method based on a low-rank-sparse information combination network, the implementation steps of which are: (1) inputting hyperspectral images; (2) obtaining low-rank information and sparse information of hyperspectral images; 3) Preprocess low-rank information and sparse information; (4) generate training set and test set; (5) construct low-rank-sparse information combination network; (6) train low-rank-sparse information combination network; (7) Classify the test samples. The invention can effectively solve the problem that the traditional low-rank recovery classification algorithm has a classification accuracy drop caused by inaccurate low-rank subspace estimation, avoids complex low-rank recovery operations, and can maintain a small number of samples while achieving high classification accuracy. recognition ability and stable classification performance.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a hyperspectral image classification method based on a low-rank-sparse information combination network in the technical field of image classification. The present invention can classify the pixels in the hyperspectral image to determine the category of ground features of each pixel, and the present invention can be applied to many fields such as disaster monitoring, geological exploration, urban planning, and target recognition. Background technique [0002] Hyperspectral records the continuous spectral characteristics of ground objects with its rich band information, and has the possibility of recognizing more types of ground objects and classifying objects with higher accuracy. However, in the process of collecting hyperspectral images, due to the error of the instrument itself, as well as the influence of climate, temperature and other factors, errors will inevi...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that due to the complex structure of the hyperspectral image, its authentic information exists in multiple low-rank subspaces. In order to accurately estimate the low-rank subspace where the authentic information of the hyperspectral image is located, it is necessary to carry out the analysis from multiple angles. Exploration, this complex operation will bring a lot of calculations, making the classification method more troublesome to implement

However, in the case of inaccurate low-rank subspace estimation, the recovered low-rank information cannot contain all the true information, and the noise cannot be completely removed.

Due to the complexity of hyperspectral images, it is very difficult to obtain low-rank information without noise. If the low-rank information with noise is used for classification, the classification effect is not ideal.

On the other hand, the neural network model trained by using a single information is usually less robust, prone to overfitting, and the classification accuracy is not high

Images