Detection And Identification Of Surface Defects Of Photovoltaic Panels Based On Industrial UAV

The installation of photovoltaic panels has a wide range of applications.Among them,large-scale photovoltaic power plants are mostly located in vast areas,and long-term exposure in the natural environment will cause varying degrees of damage to the panels.In addition,the material of photovoltaic panel and its own design and processing technology affect the service life of the panels.The power generation of damaged solar panels will be greatly reduced.In severe case,fires will occur,causing direct economic losses and endangering the lives of workers.Therefore,the inspection and monitoring of photovoltaic power stations is extremely important.In this research,industrial UAV is equipped with visible light and infrared cameras to collect images,and relevant image processing algorithms are used to process images,so as to provided data support for fault detection of largescale photovoltaic power stations.Firstly,the development status of photovoltaic industry and the application status of detection technology in recent years are analyzed.Considering that image collection site comes from large scale photovoltaic power station,the status of domestic and foreign UAV technology application is described,and UAV technology is used to for efficient image acquisition.Secondly,according to the real-time environment of the collection site and the layout of the photovoltaic modules,the patrol collection plan is formulated,the UAV patrol collection system is introduced,the preliminary safety flight correction work is completed,and the working principle of photovoltaic panel power generation is analyzed and studied to find the cause of hot spots.Then,to deal with the problems of motion blur,noise interference,poor image contrast,and poor brightness in the photovoltaic panel images collected by industrial drones,image preprocessing is performed.The inverse filtering and Wiener filtering are used for deblurring,median and Gaussian filtering are used to denoise salt and pepper noise and white noise in the image,and the method of piecewise linear transformation and histogram transformation are used for image enhancement research.And then,using the traditional image segmentation methods,such as photovoltaic panel segmentation based on edge detection operators,threshold segmentation and region segmentation,a single and multiple photovoltaic panel foreground segmentation based on RGB and HSV color spaces is proposed.A simulation system is designed for visible light images to realize the identification and classification of defect areas.Finally,for the segmentation of photovoltaic infrared images,the performance of infrared images is analyzed,and a hot spot detection method based on Unet network and HSV color space is proposed.According to infrared image features,the main reasons for image bright spots based on gray threshold classification are proposed.