结合倒置残差与多分支特征融合的PVC地板检测方法研究OA

To address the issues of complex background interference,missed detection of dense small targets,and high computational demands in PVC flooring defect detection,this paper proposes an improved YOLOv10 detection model.A PPSCSA attention module is designed to efficiently integrate global contextual information,enhancing the algorithm's multi-level feature representation capabilities.The C2f_iRPS module is introduced to effectively expand the receptive field,improving the algorithm's robustness against complex backgrounds.The EFC feature fusion module is incorporated to strengthen inter-level semantic correlations while reducing model parameters and computational complexity.The U-IoU loss function is adopted to improve detection accuracy for densely packed small targets with similar aspect ratios.Experimental results demonstrate that the improved algorithm achieves precision,recall,and mAP@0.5 scores of 79.3%,67.3%,and 71.9%,respectively,representing improvements of 2.8,2.3,and 3.5 percentage points over YOLOv10n,with a 9.3%reduction in model parameters.The method simultaneously achieves enhanced detection accuracy and real-time performance 251 FPS for PVC flooring inspection.