To address the unstable quality of processed bamboo and wood products caused by both the performance limitations of current processing machinery and the inherent defects of bamboo and wood materials, this study focuses on key defect detection technologies within intelligent wood manufacturing, and proposes an integrated quality assessment system based on an improved YOLO （You Only Look Once）object detection algorithm. Innovatively, the system incorporates a multi-parameter defect weighting mechanism, enabling quantitative analysis of critical indicators such as defect size, characteristics, and severity. Subsequently, a wood quality grading model was constructed using fuzzy comprehensive evaluation. Experimental results demonstrate that the system effectively categorizes wood products into three grades: superior (Grade A), qualified (Grade B), and unqualified (Grade C). For the Plywood wood defect dataset, the system achieved mean average precision@0.5 （mAP@0.5） of 91.3%. Moreover, the dynamic weighting-based grading strategy showed a deviation of less than 5% compared to manual evaluation results (Euclidean distance: 0.063; Jaccard index: 0.892). This research provides an efficient and scalable quality assessment paradigm for intelligent wood manufacturing, demonstrating significant engineering application value.