Improved YOLOv8n Photovoltaic Cell Defect Detection Algorithm

In order to solve the problems of inaccurate target positioning, missed detection and false detection due to factors such as difference in defect scale, complex shape and diversified defect types in photovoltaic cell defect detection, an improved YOLOv8n photovoltaic cell defect detection algorithm was proposed. Firstly, the C2F_DCNv4 module is constructed as part of the feature extraction module of the backbone and neck network, and its dynamic defect extraction characteristics and memory access optimization capabilities are used to enhance the feature extraction ability and processing speed of the network. Secondly, the ESPPM module is designed to replace the original space pyramid pooling module, which reduces the local information loss and enhances the network's ability to capture key features by recalibrating and activating channel features and global features. Finally, the CPEN (Central Penalized EIoU with NWD Loss) loss function was designed as the bounding box loss function to strengthen the target positioning and further optimize the convergence and detection accuracy of the network. The experimental results show that compared with the original mAP@0 algorithm, the improved YOLOv8n is improved by 2.1%, the computational cost is reduced by 0.3GFLOPs, the parameter quantity is reduced by 0.05M, and the detection speed reaches 128FPS, which meets the requirements of photovoltaic cell defect detection.