Machine vision based insulator image chip drop fault recognition method

Abstract

Abstract The paper studies the method of insulator image chip drop fault recognition based on machine vision to improve the recognition effect of insulator chip drop fault.Zhang’s calibration method was used to calibrate the internal parameters of binocular vision camera to improve the results in image acquisition. Insulator images were captured by binocular vision camera after calibration. The color image is converted to digital image and the insulator target image is extracted from the saturation component of a digital image by the optimal entropy threshold segmentation method. The nonlinear function is used to correct the high-frequency subband coefficients in different scales and different directions in the insulator target image, and the coefficients are used to extract the high-frequency information of different scales and directions.. The improved one-dimensional Hough transform detection algorithm is used to identify the ellipse in the edge image, that is, the insulator. The decision conditions of fault identification are designed to complete the fault identification of insulator drop. Experiments show that this method can clearly collect insulator images under different weather conditions, effectively segment insulator target images and accurately extract insulator edge information. The method can accurately identify the insulator drop fault based on the ellipse detection results. When the contrast and brightness factors are different, the f1-score value of fault recognition is higher and the fault recognition effect is better.