Power equipment post-earthquake hyperspectral image dimensionality reduction and damage classification method

Abstract

The invention provides a post-earthquake hyperspectral image dimensionality reduction and damage classification method for power equipment. The method comprises the following steps: acquiring a post-earthquake hyperspectral image of the power equipment with a determined fault type; preprocessing the collected power equipment hyperspectral image data of which the fault type is determined; reducing the dimension of the original fault hyperspectral image data set by using an LDA algorithm; dividing the data set after dimension reduction into a training set and a test set; building a GCN classifier, and using a stochastic gradient descent optimization algorithm; using the training set to train a GCN classifier; applying the test set to the trained GCN classifier to detect the classification accuracy; and inputting a post-earthquake power equipment hyperspectral image acquired by field detection into the combined model to obtain a damage classification result. According to the method, the dimension of the hyperspectral image of the post-earthquake power equipment is reduced by using the LDA algorithm, and then the hyperspectral image is classified by using the GCN classifier, so that the scientificity, high efficiency and accuracy of internal defect detection and diagnosis of the post-earthquake power equipment are effectively improved.

Description

technical field [0001] The present application relates to the technical field of internal defect identification of power equipment, and in particular to a method for dimensionality reduction and damage classification of post-earthquake hyperspectral images of power equipment. Background technique [0002] my country is a country with serious earthquake disasters. The damage caused by earthquakes to power facilities is catastrophic. The high post-disaster recovery and reconstruction costs and the huge losses caused by power outages will have a huge impact on the national economy and people's lives. . [0003] The translation invariance of the traditional CNN algorithm when processing Euclidean structure data such as images and videos is the key to its good performance in the field of computer vision, but for non-Euclidean structure data (topological maps) such as social networks, due to each The adjacent points of a vertex may be different, and translation invariance cannot b...

AI Technical Summary

Problems solved by technology

[0005] This application provides a hyperspectral image dimensionality reduction and damage classification method for power equipment after an earthquake to solve the problem of intelligent identification of internal defects of power equipment after an earthquake

Images