基于改进YOLOv5的PCB板表面缺陷检测

Surface Defect Detection of PCB Based on Improved YOLOv5

针对当前PCB板检测参数量庞大、检测精度低等问题,提出了一种改进YOLOv5的检测模型。以YOLOv5模型为框架,采用EfficientNetV2结构替换原始模型的主干网络,针对小目标缺陷,引入对空间信息更敏感的CA注意力机制,并采用α-IoU损失函数提高模型回归精度。实验结果表明:改进后的YOLOv5网络模型较原始网络均值平均精度提高了2.6%,参数量减少47%,可应用在小型工业检测设备中。

In order to solve the problems of large number of parameters and low detection accuracy of current PCB detection,a detection model of improved YOLOv5 was proposed.Taking the YOLOv5 model as the framework,the main network of the original model was replaced by the structure of EfficientNetV2.Aiming at the defects of small targets,the CA attention mechanism which is more sensitive to spatial information was introduced,and theα-IOU loss function was used to improve the accuracy of the model regression.The experimental results show that compared with the original network,the improved YOLOv5 network model improves the mean average accuracy by 2.6%and reduces the number of parameters by 47%,which can be easily deployed in small industrial testing equipment.