基于深度学习的磁环表面缺陷检测算法

Defect Detection Algorithm of Magnetic Ring Surface Based on Deep Learning

在磁环的生产制造过程中,常常由于生产环境、制造工艺等因素,难免会使磁环表面出现各种类型的缺陷。针对传统人工检测低效、耗时、检测精度低的缺点,本文提出了一种基于YOLOv3的磁环表面缺陷检测方法。实验结果表明,YOLOv3的平均识别精度达到了96.19%,单张图片检测速度达到了24.46 ms,该方法在磁环缺陷检测上有一定的先进性和有效性。

In the manufacturing process of the magnetic ring, due to the production environment and manu-facturing process factors, it is inevitable that various types of defects may appear. In view of the shortcomings of traditional manual detection, such as low efficiency, time consuming and low ac-curacy, this paper proposes a magnetic ring surface defect detection method based on YOLOv3. Experimental results show that the average recognition accuracy of YOLOv3 has reached 96.19%, and the speed of single image detection has reached 24.46 ms. This method has a certain degree of advancement and effectiveness in the detection of magnetic ring defects.