基于YOLO_v4的空心杯电枢表面缺陷实时检测

Real-time Detection of Surface Defects of Hollow Cup Armature Based on YOLO_v4

在空心杯电枢的绕制工艺过程中,绕线机异常工作会造成电枢表面出现孔洞等影响空心杯电机寿命的缺陷,为解决空心杯电枢表面微小缺陷检测过程中存在的准确率低、检测速度慢、不能实时检测缺陷等问题,文章提出一种基于YOLO_v4的空心杯电枢表面孔洞缺陷检测方法。对采集的图片进行数据增强及K折交叉验证,提高模型的鲁棒性,以避免训练模型过拟合;借助CSPDarknet53网络及SPP模块提取输入原始图像的特征,通过训练获得针对空心杯电枢表面缺陷的检测模型,提升YOLO_v4缺陷位置检测及识别的精度;在搭建的实验平台上采集数据并验证基于YOLO_v4提出方法的有效性。实验结果表明该方法可有效满足工业生产复杂背景下电机电枢表面微小孔洞缺陷检测的要求。

During the winding process of the hollow cup armature,the abnormal operation of the winding machine will cause holes on the surface of the armature and other defects that affect the life of the hollow cup motor.In order to solve the problem of small defects on the surface of the hollow cup armature,the accuracy of the detection process is low,the detection speed is slow,and the defects cannot be detected in real time.This paper proposes a method for detecting holes on the surface of the hollow cup armature based on YOLO_v4.Perform data enhancement and K-fold cross-validation on the collected pictures to improve the robustness of the model to avoid overfitting of the training model;use the CSPDarknet53 network and the SPP module to extract the features of the input original image,and obtain the surface of the hollow cup armature through training The defect detection model improves the accuracy of YOLO_v4 defect location detection and recognition;collect data on the built experimental platform and verify the effectiveness of the proposed method based on YOLO_v4.Experimental results show that this method can effectively meet the requirements of detecting micro-hole defects on the surface of the motor armature under the complex background of industrial production.