Real-Time Defect Detection of Photovoltaic Cells Based on Lightweight YOLOv8

Along with the rapid development of solar energy, photovoltaic (PV) cell defect detection has gradually become a global industrial issue and research hotspot of great concern. Ways to enhance detection precision, lower the rate of false positives and improve the system efficiency are the challenges ongoing in the domain of solar cell fault detection. As a result, an enhanced YOLOv8-based approach for detecting cell defects is presented. In order to increase the inference speed on CPU, a lightweight network PaddlePaddle-Lightweight CPU Net (PPLCNet) is used to serve as the core architecture. Alongside adopting the MPDIoU metric for spatial similarity assessment based on extremal point distances, the system introduces a reconstructed loss function that optimizes bounding box parameter regression through geometric regularization. Experiments show that this method achieves a 3.2 % increase in detection accuracy compared to YOLOv8, the detection time of a single image is only 0.6 ms, and the number of model parameters is 5. 9 M. This method shows excellent detection speed and low computing requirements in photovoltaic cell defect detection, and is especially suitable for the application scenarios of real-time monitoring and efficient processing, meeting the needs of fast response and efficient calculation.