CHS-YOLO:Enhanced Lightweight YOLOv11 Model for Accurate Photovoltaic Panel Defect Detection

In order to solve the performance bottleneck and background interference problems of traditional PV panel defect detection algorithms, this paper proposes an improved framework, CHS-YOLOv11, whose core innovations include: the C3k2_DEConv module enhances multi-scale defect sensing capability; the HSPAN necking network integrates multi-scale spatial attention and hierarchical feature reorganization; the Shape-IoU loss function introduces shape consistency constraints to optimize the bounding box regression accuracy and shape fitting, which significantly improves the effectiveness of tiny crack and irregularity detection. Optimize the bounding box regression accuracy and shape fitting to significantly improve the detection effect of tiny cracks and irregular defects. Experiments on a photovoltaic panel defect dataset demonstrate that CHS-YOLOv11 outperforms the original YOLOv11 in key metrics such as average precision (mAP@0.5 improved by 4.3%), recall (R improved by 5.3%), and precision (P improved by 3.7%), with a 26.9% reduction in model parameters. This study offers an efficient and practical solution for lightweight photovoltaic panel defect detection.