基于动态蛇形卷积的轻量化PCB缺陷检测算法OA

In view of the problems of insufficient detection accuracy and large number of parameters in the ex-isting PCB(Printed Circuit Board)defect detection algorithms,a high-precision lightweight detection algorithm YO-LO-DLN(YOLO-Dynamic Lightweight Network)is proposed based on YOLOv8(You Only Look Once version 8).Dynamic serpentine convolution is introduced into the backbone network to capture more fine structural features of the PCB surface,and the extracted features are processed by the lightweight detection head LSCDHead(Lightweight Shared Convolutional Detection Head).Features of different scales share the same set of convolution to reduce the number of model parameters and perform group normalization operations.The NWD(Normalized Wasserstein Dis-tance)loss function with low sensitivity to the target scale is utilized to evaluate the bounding box similarity,further improving the detection accuracy of the proposed method.In the experiment,YOLO-DLN increases mAP50(mean Average Precision50)and MAP50:95 by 4.4%and 4.6%respectively while using 88.1%of the parameters of the original model.The experimental results show that YOLO-DLN can maintain a high detection accuracy while signifi-cantly reducing the number of model parameters,and is suitable for PCB defect detection in resource-constrained en-vironments.