Research On Large-Scale Photovoltaic Panel Detection Technology Based On UAV Vision

As the photovoltaic panels are usually placed in the harsh outdoor environment,perennial withstand wind and sun,resulting in its serious problem of malfunctions and defects.It is an effective scheme to meet the requirements of the extensive,accurate and real-time requirements of large-scale PV panel defect detection system by adopting the machine vision technology and image processing technology to process the aerial images of PV panel,which collected by the UAV.Therefore,it is very important to study and experiment the method of image preprocessing,image stitching and image recognition of the images collected by the UAV for the defects detection of large-scale photovoltaic panels.Firstly,the methods of image preprocessing and image correction are studied for the problem of noise interference and geometrical deformation of the PV panel images collected by UAV.The image preprocessing part adopts the weighted average image grayation method,the median image filtering method and the maximum interclass variance image binarization method.And the image correction part adopts the tilt correction method based on the Hough transform and the perspective correction method based on connection points.Experiments show that the image preprocessing method and the correction method could effectively solve the problem of noise interference and geometric deformation of the aerial images of the PV panel,and then obtained the PV panel images which were suitable for the subsequent processing.Secondly,in order to stitch the aerial images of several large-scale photovoltaic panels into a complete image,an improved SURF-BRISK aerial image stitching method is studied.The improved SURF feature detection algorithm,the BRISK feature description algorithm,the Hamming distance feature rough matching method,the RANSAC feature fine matching method,the bilinear interpolation method and the weighted fusion method are used to realize the image stitching of the aerial images of the PV panel.Experiments show that the image stitching method used in this paper could effectively obtain the complete image of PV panel.Moreover,compared with the image stitching methods which are commonly used,this method had obvious advantages in terms of real-time and accuracy.Finally,for the problem of defects such as crack in PV panel,a method of detecting defects of PV panel based on image feature is studied.This method realizes the distinction and detection of different defects,such as crack,broken corner,broken pole,broken grid,dirty,etc.,by analyzing the characteristics of image feature,such as area,length and width,rectangle and direction.And the crack defect of the PV panel was studied and analyzed,to verify the validity of the defect detection method used in this paper.Moreover,it provides useful information for the maintenance and replacement work of the PV panel,in order to improve the efficiency of photovoltaic power plant.