To address the challenges caused by diverse photovoltaic panel surface states categories, uneven target-scale distributions among different categories, and complex appearance variations such as dust coverage, snow coverage, and physical damage, a photovoltaic panel surface statea defect detection method based on YOLOv11-CND4H is proposed. By introducing the Coordinate Attention mechanism, the network''s perception of key defective regions, occluded regions, and spatial positional information is enhanced.the network’s response to key regions and fine-grained defect features is enhanced, while feature interference caused by complex backgrounds is suppressed. A four-scale decoupled detection head based on Diverse Branch Block re-parameterization is further incorporated to strengthen multi-scale feature representation and fusion, thereby improving the model''s ability to represent photovoltaic panel surface states targets with different scales and morphological characteristics. detection performance for small-sized defect targets. In addition, a collaborative localization loss combining Normalized Wasserstein Distance and Complete Intersection over Union is employed to optimize the bounding-box regression process and further improve localization accuracy. Experimental results show that, compared with the baseline YOLOv11 model, the proposed method improves precision by 3.97%, while increasing mAP50 and mAP50-95 by 3.23% and 2.26%, respectively. The proposed method significantly outperforms existing mainstream object detection algorithms, effectively improving the detection accuracy and boundary localization quality of small and weak-textured defects multi-category photovoltaic panel surface states targets and defectis onon photovoltaic panels under complex backgrounds, and demonstrates promising engineering application value.