With the growing demand for wooden furniture, accurate and efficient detection of edge banding defects in wood panels has become increasingly important. To address the limitations of existing methods—such as low accuracy, high missed detection rates, and frequent false positives—this study proposes an improved YOLOv8-based detection algorithm, termed YOLO-MDEW. Built upon the YOLOv8n framework, the model integrates several key enhancements: the C2f-MCFF(Multi-Channel Feature Fusion) module replaces the original C2f structure to improve multi-scale feature extraction; an enhanced SPPF-D(Spatial Pyramid Pooling with Dilation) module is incorporated to strengthen cross-scale information fusion; and an Efficient Local Attention (ELA) mechanism is applied within SPPF-D to better capture fine-grained defect features. Additionally, the original CIoU loss is replaced with Wise-IoU v3 (WIoU) to accelerate convergence and improve localization accuracy. Experimental results on a custom-built wood panel edge defect dataset demonstrate that YOLO-MDEW achieves a mean average precision (mAP) of 74.0%, representing a 1.9% improvement over the baseline YOLOv8n. These results highlight the proposed method’s enhanced robustness and effectiveness in detecting edge banding defects in wooden panels.
Xiao et al. (Fri,) studied this question.