Adaptive Classification Method for Multispectral Remote Sensing Images Based on Spectral Labeling

Abstract

The invention discloses a spectral tag based adaptive multi-spectral remote sensing image classification method, and mainly solves the problem that automatic classification is difficult to realize due to the fact that manual operation is required during remote sensing image classification in the prior art. The method includes the realization steps: (1) inputting multi-spectral remote sensing images; (2) adjusting data formats of the multi-spectral remote sensing images; (3) completing a spectral tag library through the adjusted multi-spectral remote sensing images; (4) adjusting spectral tags through a K-Means clustering algorithm; (5) adopting the adjusted spectral tags as weak supervisory information to perform weak supervision classification to obtain classification results. By the method, manual operation during classification is avoided, classification precision is improved, classification speed is increased, and the method can be used for land coverage information analysis.

Description

technical field [0001] The invention belongs to the field of image processing, and further relates to a method for classifying multispectral remote sensing images, which can be used for analyzing land cover conditions. Background technique [0002] Multi-spectral remote sensing images are high-resolution images obtained by scanning the ground with satellite multi-spectral scanning systems, and their rich information provides the possibility for computer recognition and classification of ground object images. However, due to its low spectral dimension, the extraction of land use and coverage information has become one of the difficulties in remote sensing information processing. With the rapid development of the national economy, the degree and speed of changes in land use are accelerating, and there is an urgent need for fast and accurate land use information acquisition methods, that is, to classify remote sensing images according to land use. [0003] There are many exist...

AI Technical Summary

Problems solved by technology

[0008] In summary, the existing multispectral remote sensing image classification methods have the disadvantages of low classification accuracy, slow execution speed, and poor universal applicability of training data. At the same time, the above classification methods need to manually set the number of classifications, making it difficult to achieve automatic classification.

Images