Defect Detection in Computer Motherboard Assembly through Fusion of Multi-Scale Features and Attention Mechanisms

To meet the growing demands for automation and intelligence in the quality inspection of computer motherboard assembly, this study proposes an improved defect detection method based on RetinaNet. The method integrates multi-scale features and attention mechanisms. A path aggregation network is introduced to construct a feature fusion network, combining low-level positional features with high-level semantic information to capture features of different scales in target objects. Furthermore, an attention mechanism is introduced after feature fusion to weight feature maps, enhancing the representation capability of foreground defect targets and mitigating interference challenges in the motherboard defect detection process. Experimental validation on a self-built dataset for computer motherboard assembly defect detection demonstrates an average detection accuracy of 84.5%. The results underscores the effectiveness of the proposed method, providing a promising solution for the automation and intelligence enhancement of assembly inspection tasks.