Method for positioning insulator in image

Abstract

The invention discloses a method for positioning an insulator in an image, and relates to the field of power equipment image target positioning. Before pollution, crack, burst loss and other external defects of insulators of overhead lines and towers are detected, insulators in acquired images of devices such as unmanned aerial vehicles need to be accurately positioned. The method comprises the following steps: 1) acquiring an insulator image; 2) expanding the image; 3) labeling the positions of the insulators in the acquired image and the expanded image by using a labelImg tool; 4) constructing an image training set and a test set; 5) performing insulator target model detection training based on an improved SSD algorithm; 6) testing the trained model system; and 7) after the test is passed, inputting an actually acquired insulator image for positioning and distinguishing. Accurate positioning and identification of the insulator image can be conveniently realized, and possibility is provided for further insulator appearance defect detection.

Description

technical field [0001] The invention relates to the field of image target positioning of electric equipment, in particular to a method for locating an insulator in an image. Background technique [0002] Insulator A piece of electrical equipment consisting of glass or ceramic used to increase creepage distances, achieve electrical insulation and mechanical fixation. Because the insulator is exposed to the outside world for a long time, it is affected by external force damage such as hail and other bad weather, overvoltage, pollution flashover, etc., causing the insulator to drop strings or self-explode, unable to complete mechanical support and electrical insulation, resulting in power grid safety accidents. Therefore, it is necessary to detect the external defects of insulators such as contamination, cracks, and missing bursts. [0003] Before detection, it is first necessary to accurately locate the insulators in the images collected by drones and other devices. Existing...

AI Technical Summary

Problems solved by technology

Existing target localization algorithms based on deep learning generally complete model training by learning a large number of images, but it is difficult to collect a large number of high-quality insulator images in actual model training

In addition, the original SSD algorithm does not fully consider the context information, resulting in insufficient information for target detection in each scale, and the detection accuracy is not high enough

Images