Research On Defect Detection And Location Technology Of Photovoltaic Panel Based On Vision

In recent years,China’s solar power generation technology has developed rapidly,and the scale of photovoltaic power generation is also doubling.Most of China’s photovoltaic electric fields are distributed in the northwest and northeast.In these areas,there is little precipitation,heavy wind and sand,and much dust.Dust and dirt are easy to accumulate on the surface of photovoltaic panels.At the same time,these stains will cover the surface of photovoltaic panels,resulting in the failure of photovoltaic panels due to "hot spot effect",Manual cleaning and detection of photovoltaic panel defects requires a lot of manpower and material resources.Therefore,it is necessary to detect photovoltaic panel defects independently through photovoltaic cleaning robot.This paper mainly studies the software of photovoltaic cleaning robot.The main work is as follows:(1)Using the traditional pixel operation method to detect the defect image of photovoltaic panel.According to the characteristics of photovoltaic panel defect images,this paper designs a set of detection algorithm based on pixel operation,including key technologies such as HSV space conversion and connected domain label detection,and randomly selects 200 photovoltaic panel defect images for detection,which verifies the accuracy and efficiency of the algorithm.(2)Using the method based on deep learning to detect the defect image of photovoltaic panel.In this paper,Faster R-CNN network and YOLOv3 network are used for training and detection respectively.After comparing the accuracy,recall and other parameters of the two methods,YOLOv3 network is selected to improve from three aspects,including using adaptive learning rate configuration method,K-means clustering algorithm and lightweight model Tiny-YOLOv3,and then experiments are carried out with the improved network.The experimental results show that the network has better accuracy and efficiency than the traditional pixel operation method and the first two deep learning networks.(3)Detecting the infrared image of photovoltaic panels with defects inside.After selecting the appropriate infrared image acquisition equipment,based on the characteristics of photovoltaic panel infrared image,a set of algorithm for detecting photovoltaic panel internal defects based on threshold segmentation is designed,and the effectiveness of the algorithm is verified by experiments.(4)Studying the defect location technology of photovoltaic panels.In this paper,the position of the robot on the photovoltaic panel is obtained by the relative positioning method based on the encoder,and then combined with the spatial coordinate conversion and camera calibration,the camera internal parameters and distortion parameters are obtained.Then,the actual distance from the defect to the robot is obtained by using the single frame still image model,so as to realize the defect positioning.