An intelligent defect detection method for printed circuit boards based on optimized YOLOv8-X
To improve the accuracy and efficiency of printed circuit board(PCB)defect detection,this paper proposes an opti-mized YOLOv8-X-based intelligent detection method.By systematically optimizing the network structure,activation functions,and loss functions,the model's performance is significantly enhanced.First,the CBAM attention mechanism is introduced in the back-bone layer to strengthen feature correlation.Then,the traditional convolution modules in the neck network are replaced with RepNCSPELAN4,improving the model's representational capacity.Furthermore,the loss function in the head network is replaced with Generalized IoU,effectively addressing small object detection and class imbalance issues,thus enhancing the model's ro-bustness.Finally,Leaky ReLU is employed in place of ReLU as the activation function,improving the model's nonlinear feature representation,making it better suited for complex defect detection scenarios.Experimental results demonstrate that the opti-mized YOLOv8-X model achieves significant improvements in accuracy and robustness for PCB defect detection tasks,showcasing its broad application potential in industrial inspection.
万方数据
