Based on the improved YOLOv8 photovoltaic module defect detection method

The photovoltaic power station covers a wide area, the detection environment is complex, and the defect recognition of photovoltaic modules is difficult. An improved defect recognition and detection method of photovoltaic modules based on YOLOv8 model is proposed. Based on the backbone network structure, the progressive feature pyramid (AFPN) is used to fuse images at different levels to extract multi-scale information. Instead of each pooling layer and each step convolution layer of the original model, a SPD convolution neural network is established to improve the defect recognition ability of small target features such as hot spots and black edges in photovoltaic panel images. The improved YOLOv8 model improves the fault recognition ability of small target features such as hot spots and black edges in photovoltaic module images. The experimental results show that the recall rate and accuracy of the improved model are better than those of the original model, and the average accuracy reaches $\mathbf{8 4. 2 \%}$.The detection and recognition of small target defects of photovoltaic cells are realized, and the feasibility of the model is verified.