UAV-based Binocular Vision For Insulator Fault Detection

In recent years,China’s electric power industry has developed rapidly,and a large number of transmission lines has been laid.As an indispensable electrical component in transmission lines,insulators,which are exposed to the field for a long time,can become a safety hazard for electrical energy transmission due to their frequent faults such as spontaneous explosion and cracking.With the maturity of UAV technology,UAVs are widely used in the field of insulator fault detection,because of its high efficiency,portability,low cost and remote control.Currently,insulator images are commonly acquired in a monocular way,but the monocular cameras have a limited range and cannot capture the full target,which can lead to a decrease in the accuracy of fault detection.Therefore,this paper conducts a study based on binocular vision technology and image processing technology to finally complete the detection of insulator faults.This paper firstly introduces the basic principles of binocular vision,the imaging model and the conversion between coordinate systems in detail,and applies the camera calibration method to realise the calibration of the vision system and obtain the calibration parameters of the camera in order to improve the accuracy of the binocular camera imaging system.Secondly,the key techniques of image pre-processing are introduced to address the problem of redundant information in colour images by greyscaling them,effectively reducing the amount of operations and equalising the greyscale maps to avoid uneven brightness in the images.In order to eliminate noise interference in the image,noise reduction is applied to the image using median filtering.When segmenting the insulator strings from the background,a maximum inter-class variance segmentation algorithm based on least squares fitting is used to increase the computational speed of the system while ensuring the segmentation effect.Again,in the study of stereo matching of binocular images,the BRISK feature descriptor is introduced to optimise the ORB feature matching algorithm,which not only improves the scale invariance but also significantly reduces the amount of operations,based on which the wavelet transform algorithm is used to realise the fusion of insulator binocular images.Finally,the pre-processed aerial images are used for insulator fault detection,focusing on the "self-burst" fault among the common insulator faults,using an improved Hough transform method to obtain insulator ellipse parameter information,and designing a classification algorithm based on the ellipse parameters to detect insulator strings,meanwhile,the surface crack and pollution fault of insulator are also analysed.In the simulation test of image segmentation,the algorithm in this paper was able to reduce the number of calculations significantly,saving about 50% of the system running time.In the image stereo matching simulation,the optimised ORB algorithm increased the correct matching rate from 78.6% to 96.5%.The fault detection algorithm in this paper was able to accurately segment the insulator contours in the aerial images of the UAV and diagnose the fault condition of the insulators.After simulating and analyzing 200 insulator images,38 fault images were successfully detected,2 were missed,and the recognition rate of self-exploding faults could reach95%,proving that the algorithm used in this paper has some reference value for intelligent power line inspection.