Abstract Printed Circuit Boards (PCB) are the core components of electronic devices, and their quality directly affects product performance and reliability. With increasing industrial automation, printed circuit board (PCB) defect detection technology has become crucial for ensuring production quality and efficiency. However, traditional detection methods have limitations in terms of detection accuracy, adaptability to complex backgrounds, and computational efficiency. Therefore, this study proposes a target detection model based on an improved YOLOv11 algorithm, named DefectFusionNet. By incorporating the Cross-Stage Partial DualBlock (CSP-DualBlock) for enhanced multi-scale feature extraction, the Flexible Dual-path Scale Network-Dynamic Adaptive Scale Fusion (FDSN-DASF) for dynamic adaptive feature fusion, DeepDown for improved small target retention during downsampling, and Inner-Multi-Point Directional Intersection over Union (Inner-MPDIoU) for optimized loss function design, the model significantly improves both detection accuracy and computational efficiency in PCB defect detection tasks. The experimental results show that, compared to the baseline model, the proposed model achieves a 2.6% increase in mean Average Precision (mAP) 50 and a 13.7% increase in mAP@50-95, demonstrating high efficiency and accuracy in practical measurements. The proposed DefectFusionNet demonstrates superior performance in detecting small-scale defects and handling complex background scenarios compared with existing YOLO-based approaches.
Liu et al. (Thu,) studied this question.