基于YOLOv4改进算法的零件表面缺陷检测  

Surface Defect Detection of Parts Based on YOLOv4 Improved Algorithm

作　　者：王宪伦[1] 孙宇轩 王栋 WANG Xianlun;SUN Yuxuan;WANG Dong(School of Mechanical and Electrical Engineering,Qingdao University of Science and Technology,Qingdao 266100)

机构地区：[1]青岛科技大学机电工程学院,青岛266100

出　　处：《计算机与数字工程》2023年第2期520-525,共6页Computer & Digital Engineering

摘　　要：在零件加工过程中,由于工况多样性,导致零件表面缺陷繁多,并且检测起来准确率低。针对零件表面缺陷检测问题,提出了基于YOLOv4改进算法的零件表面缺陷检测系统。论文采用了德国DAGM2007数据集作为缺陷样本,采用数据增强的方式对图片数量进行扩充。采用稠密网络模块对YOLOv4算法进行改进。最后利用改进的YOLOv4算法进行缺陷检测定位,准确率达到97%。与原算法对比,改进算法的检测速度明显提升,具有较好的应用价值。In the process of parts processing,due to the diversity of working conditions,the surface defects of parts are numerous,and the detection accuracy is low.Aiming at the problem of surface defect detection of parts,a surface defect detection system based on YOLOv4 improved algorithm is proposed.In this paper,the German DAGM2007 data set is used as the defect sample to expand the number of images by means of data enhancement.Using dense network module to improve YOLOV4 algorithm.Finally,the improved YOLOv4 algorithm is used to detect and locate defects,and the accuracy rate reaches 97%.It has good application value.

关 键 词：表面缺陷检测 YOLOv4 神经网络 数据增强 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]