YOLOv8-BGC:基于YOLOv8改进的光伏电池缺陷检测方法OA

Electroluminescence(EL)imaging is an effective method for detecting photovoltaic(PV)modules.The high spatial resolution pro-vided by EL images can detect the slightest defects on the surface of PV modules.However,the analysis of EL images is usually a manual pro-cess that is costly,time-consuming,and requires specialized knowledge of different types of defects.Therefore,a lightweight photovoltaic cell defect detection algorithm YOLOv8-BGC is proposed to improve YOLOv8.Firstly,combining the advantages of CNN and Transformer,a BOT module is proposed that can extract global and local feature information of photovoltaic cell defect images to adapt to the defect features of pho-tovoltaic cells;Secondly,coordinate attention mechanism(GAM)is introduced at the end of the backbone network and the neck network to re-duce information fragmentation and enhance global dimensional feature interaction,thereby improving the operational efficiency,feature de-scription,and parsing ability of the model;Finally,the C2fGhost module is used in the YOLOv8 neck network to reduce floating-point opera-tions during feature channel fusion,decrease model parameter count,and improve feature expression performance.Experiments have shown that on datasets with different defects in photovoltaic cells,the improved YOLOv8-BGC reduces model parameters and computational com-plexity by 6.7%and 8.5%respectively compared to the original YOLOv8 model,and increases mAP50 by 3%.It improves model accuracy and real-time performance while being lightweight.Compared with other algorithms,it also has certain advantages and can meet industrial deploy-ment requirements.