PHOTOVOLTAIC DEFECT DETECTION BASED ON IMPROVED YOLOv5 ALGORITHM

Aiming at the limitations of the previous PV defect detection，such as fewer types of recognizable defects，multiple model parameters，large volume of model parameters，and slow detection speed，the traditional YOLOv5 network is improved to detect and classify the five main types of defects，namely，cracks，broken grids，black cores，thick wires，and hot spots that are commonly found in the images of photovoltaic panels. Three different attention mechanism modules：CA attention mechanism module，ECA attention mechanism module，and CBAM attention mechanism module，are integrated into the YOLOv5 network for comparative analysis experiments，and it is found that the CA attention mechanism is more suitable for PV defect image detection. Subsequently，the YOLOv5 algorithm incorporating the CA attention mechanism module is added to the bidirectional feature pyramid network structure further to strengthen the feature fusion capability of the network. The experimental results show that the model can effectively identify and localize five types of common defects，and compared with the YOLOv5 algorithm，the Map（Mean Average Precision）value is improved by 3.7 %，the model volume is reduced by 15 %，and the average speed of the detection of the images is improved by 9.7 %. The overall conclusion shows that the method effectively enhances the ability of YOLOv5 algorithm in PV defect detection，and at the same time reduces the misdetection and omission of the deep learning algorithm in PV detection. © 2024 Science Press. All rights reserved.