基于YOLOv4卷积神经网络的钢材缺陷检测系统  

STEEL DEFECT DETECTION SYSTEM BASED ON YOLOV 4 CONVOLUTIONAL NEURAL NETWORK

作　　者：黄如兵[1] 赵涟漪[2] HUANG Ru-bing;ZHAO Lian-yi(School of Information Engineering,Anhui Vocational and Technical College,Hefei 230011,China;School of Mechanical Engineering,Anhui Water Resources and Hydropower Vocational and Technical College,Hefei 231603,China)

机构地区：[1]安徽职业技术学院信息工程学院,安徽合肥230011 [2]安徽水利水电职业技术学院机械工程学院,安徽合肥231603

出　　处：《南阳理工学院学报》2022年第4期83-87,98,共6页Journal of Nanyang Institute of Technology

基　　金：2022年高校优秀青年人才支持计划项目(gxyq2022251);安徽省高校自然科学研究一般项目(KJ2018B0001);安徽高校自然科学研究项目重点项目(KJ2020A1040)。

摘　　要：为了提升钢材生产质量,设计了一种基于YOLOv4卷积神经网络的钢材缺陷检测系统。钢材缺陷目标检测系统硬件结构主要包括工控机服务器、电动控制模块、工业相机和网络通信装置,软件结构采用YOLOv4目标检测算法模型实现钢材缺陷识别、定位和筛选,产品缺陷主要包括:斑块、点蚀表面、内含物和划痕等缺陷。通过对钢材缺陷检测系统的测试验证表明,系统能根据缺陷特征准确快速地检测识别和定位缺陷目标,能满足企业生产中对钢材缺陷检测的实际需求,在智能制造领域具有一定的应用前景。In order to improve the quality of steel production,a steel defect detection system based on YOLOv4 convolutional neural network is designed.The hardware structure of the steel defect target detection system mainly includes an industrial computer server,an electric control module,an industrial camera and a network communication device.The software structure adopts the YOLOv4 target detection algorithm model to realize the identification,positioning and screening of steel defects.The product defects mainly include:plaque,Defects such as pitting surfaces,inclusions and scratches.The test and verification of the steel defect detection system shows that the system can accurately and quickly detect,identify and locate defect targets according to defect characteristics,which can meet the actual needs of steel defect detection in the production of enterprises,and has certain application prospects in the field of intelligent manufacturing.

关 键 词：YOLOv4 卷积神经网络 缺陷检测 工业相机 钢材 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]