Hyperspectral Image Classification Method Based on Spatial Information Migration

Abstract

The invention discloses a hyperspectral image classification method based on spatial information migration, which mainly solves the problems of inaccurate, slow speed and poor stability of the existing hyperspectral image classification based on the SVM algorithm. The implementation steps are: first randomly select some bands from the hyperspectral image as the source domain, use the EM algorithm to cluster the source domain to obtain spatial information; randomly select some bands in the remaining bands of the hyperspectral image as the target domain, and Under the constraints of information, the target domain samples are migrated to the labeled samples; finally, the SVM algorithm is used to train and classify these labeled samples to obtain the final classification results. Compared with the traditional classification method, the present invention has the advantages of fast speed and good effect, and can use fewer bands to achieve higher clustering accuracy, which greatly saves calculation costs and can be used for mineral exploration, resource investigation and environmental monitor.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to the processing of hyperspectral remote sensing images, and can be used for mineral detection, resource investigation and environmental monitoring. Background technique [0002] With the development of remote sensing technology and the advancement of imaging equipment, optical remote sensing technology has entered the era of hyperspectral remote sensing. The emergence and development of hyperspectral remote sensing technology will make remote sensing imaging equipment faster, higher The amount of information provides humans with massive observation data, and brings people's ability to observe and understand ground objects through remote sensing technology into a new stage. For the hyperspectral remote sensing data collected by the imaging spectrometer, how to make full and effective use of the massive information contained in it poses challenges and brings opportunities to the ...

AI Technical Summary

Problems solved by technology

Some commonly used supervised classification algorithms, such as SVM algorithm and Bayesian algorithm, can achieve good classification results when there are sufficient training samples, but due to the large number of hyperspectral image bands, the classification speed is slow, and the labeled samples Sufficiency has a great influence, and often requires a large number of labeled samples, otherwise the classification effect is not ideal

Images