Semi-supervised integrated remote-sensing image classification method based on attractor propagation clustering

Abstract

The invention discloses a semi-supervised integrated remote-sensing image classification method based on attractor propagation clustering, relates to the remote-sensing image classification field, and aims at solving the problem that the overall classification accuracy is low due to the fact that samples are not sufficient and not representative in a remote-sensing image classification process. The semi-supervised integrated remote-sensing image classification method comprises the following steps: step A, an attractor propagation clustering algorithm is utilized for controlling a self-training semi-supervised algorithm to generate individual classifiers; step B, a weighted voting method is utilized for integrating the multiple individual classifiers generated in the step A; step C, remote-sensing images are classified through a semi-supervised integrated model obtained in the step B. The semi-supervised integrated remote-sensing image classification method based on the attractor propagation clustering can be widely applied to remote-sensing image classification methods.

Description

technical field [0001] The invention relates to the field of remote sensing image classification. Background technique [0002] The statistical distribution of remote sensing image information is highly complex and random; when selecting samples, there are also factors such as limitations and blindness in the cognition of classified images. These lead to a small number of samples and are not representative enough. For small-sample problems, scholars generally use two learning paradigms, semi-supervised learning and ensemble learning, to improve classifiers. Semi-supervised learning uses the structural information of the feature type hidden in the unlabeled samples to fit a more representative classifier; integrated learning combines multiple homogeneous or heterogeneous learning machines for the same problem To learn and improve the generalization ability of the classifier. However, these two approaches have developed almost in parallel, and a recent study showed that the...

AI Technical Summary

Problems solved by technology

[0003] Aiming at the problem of low overall classification accuracy due to insufficient samples and poor representation in the remote sensing image classification process, the present invention proposes a semi-supervised integrated remote sensing image classification method based on attractor propagation clustering, with the purpose of further improving remote sensing Image classification accuracy

Images