改进Yolo V3算法在工件缺陷检测中的应用

为快速、准确地检测工业生产中工件表面产生的缺陷,提出了一种基于Yolo V3的工件表面缺陷检测方法。该方法以DarkNet卷积模型作为特征提取网络,通过引入数据增强方法防止产生过拟合现象,并针对工件表面缺陷形状单一、缺陷尺寸普遍偏小的特点改进了Yolo V3网络的特征融合方式,减少了冗余候选框的数量,提升了算法性能。以环形工件作为检测对象搭建了实验平台。实验结果表明,所提方法能克服人工提取特征的局限性,检测精度和检测速度均满足实际生产要求。

Research on the Detection of Workpiece Surface Defects Based on Yolo V3

In order to quickly and accurately detect the defects on the surface of workpieces in industrial production,a method based on Yolo V3 is proposed.In this method,DarkNet convolution model is used as feature extraction network,and data enhancement method is introduced to prevent over fitting.The feature fusion method of Yolo V3 network is improved to reduce the number of redundant candidate boxes and improve the performance of the algorithm according to the characteristics of single surface defect shape and small defect size.The experiment platform is built with the ring workpiece as the test object.The experimental results show that the proposed method can overcome the limitations of manual feature extraction,and the detection accuracy and speed meet the actual production requirements.