Hyperspectral image recognition method based on spatial clustering information correction

Abstract

The invention discloses a hyperspectral image recognition method based on spatial clustering information correction. The hyperspectral image recognition method comprises the following steps: correcting an aerial hyperspectral image; marking a part of shot hyperspectral images as a training sample set, and taking the rest as a test sample set; acquiring a set number of spectral lines from the hyperspectral images of the training sample set for each type of articles, and extracting characteristic spectral segments of the spectral lines of each type of articles by adopting a competitive adaptivereweighting algorithm; training an SVM classifier according to the characteristic spectrum segments, then extracting the characteristic spectrum segments corresponding to spectral lines in the hyperspectral images in the test sample set according to the characteristic spectrum segments, and inputting the characteristic spectrum segments into an SVM model to obtain primary classification of articles in the hyperspectral images in the test sample set; marking high-credibility pixel points in the hyperspectral image of the test sample set, and correcting the pixel points in all windows of the hyperspectral image by adopting a small window model according to the marked high-credibility pixel points to obtain updated article classification.

Description

technical field [0001] The invention relates to a method for classifying objects in an image, in particular to a hyperspectral image recognition method based on spatial clustering information correction. Background technique [0002] With the development of the country's economy, electricity consumption is increasing year by year, and the power grid has put forward greater requirements for operation safety. The development of the west-to-east power transmission project has increased the installed capacity in the southwest region year by year. Due to the many mountains in the southwest region, the transmission corridors are full of trees, and the threat of lush trees to transmission lines has gradually become one of the main threats to the safe operation of the power grid. [0003] The traditional way to prevent threats from trees is mostly manual line inspection, which has a long cycle, high risk and heavy labor. With the introduction of modern information technology and ad...

AI Technical Summary

Problems solved by technology

[0004] Nevertheless, studies have shown that the differences in the growth rates of different trees lead to certain safety hazards in the unified felling of indiscriminate trees, and the unified increase in felling will undoubtedly cause a waste of manpower

Images