Hyperspectral image classification algorithm

Abstract

The invention discloses a hyperspectral image classification algorithm, which combines a self-training semi-supervised classification algorithm and an attractor propagation algorithm, is different from the traditional self-training semi-supervised classification algorithm, and is improved in three aspects. The invention comprises the following steps: firstly, proposing the algorithm to introduce an attractor propagation algorithm to self-adaptively search an unmarked sample with high credibility; and secondly, proposing the algorithm to limit the search range of unmarked samples by using a neighborhood to improve reliability. Compared with a traditional spatial information extraction algorithm based on segmentation, the calculation amount can be effectively reduced. The invention also comprises the following steps of thirdly, proposing the algorithm to construct a similarity matrix of an attractor propagation algorithm by using a spectral correlation angle; According to the invention,the proposed algorithm is compared with a Gaussian mixture model semi-supervised classification algorithm, a Laplace support vector machine and a k-nearest neighbor-based self-training semi-supervisedclassification algorithm on a classic hyperspectral image India Pines. Compared with a comparison algorithm, the proposed algorithm has higher global classification precision and higher convergence speed.

Description

technical field [0001] The invention relates to the technical field of hyperspectral image classification, in particular to a hyperspectral image classification algorithm. Background technique [0002] The rich spatial information and spectral information of hyperspectral remote sensing images provide favorable conditions for accurate identification of ground objects. However, due to the scarcity and difficulty in obtaining labeled samples of hyperspectral data, the Hughes phenomenon is prone to occur in hyperspectral data processing. In order to solve this problem, many scholars focus on the study of semi-supervised learning algorithms. The semi-supervised learning algorithm is a kind of classification method that uses a small number of labeled samples and a large number of unlabeled samples to improve the classification accuracy under the condition of few labeled samples through a semi-supervised learning strategy. At present, the learning strategies of semi-supervised l...

AI Technical Summary

Problems solved by technology

Currently available literature mostly uses the attractor propagation algorithm combined with the feature extraction algorithm or supervised classification algorithm to improve the performance of the target algorithm, and its application in semi-supervised classification is less

Images