基于图像背景差影的冲压板料表面缺陷检测  

Machine Vision Inspection of Surface Defects of Stamping Blank

作　　者：商业彤 刘乘昊 王晶 孔祥志 何雨阳 魏鸿磊 SHANG Ye-tong;LIU Cheng-hao;WANG Jing;KONG Xiang-zhi;HE Yu-yang;WEI Hong-lei(School of Mechanical Engineering and Automation,Dalian Polytechnic University,Dalian 116034,China)

机构地区：[1]大连工业大学机械工程与自动化学院,大连116034

出　　处：《组合机床与自动化加工技术》2022年第12期85-88,共4页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：辽宁省教育厅2021年度科学研究经费面上项目(LJKZ0535,LJKZ0526);2021年度本科教育教学综合改革项(JGLX2021020,JCLX2021008)。

摘　　要：为避免毛坯板料冲压成成品之后处理困难的问题,需要对冲压毛坯板料表面缺陷进行检测。提出一种高斯金字塔加速的NCC模板匹配和背景差影法相结合的冲压毛坯板料表面缺陷检测方法,首先通过高斯金字塔结合归一化互相关(NCC)函数对待检测毛坯件匹配定位,然后利用快速傅里叶变换在频域处理图像获得无缺陷背景图像,通过背景差影法获得缺陷差分图像再利用阈值分割提取缺陷。实验证明缺陷检测准确率高达96.7%,每幅图像检测耗时低于0.054 s,满足工业生产速度要求。In order to avoid the difficult problem of processing blank sheet metal after stamping into finished products,it is necessary to detect the surface defects of blank sheet metal.A gaussian pyramid acceleration of NCC template matching method and background difference method stamping blank sheet metal surface defect detection method,first by gaussian pyramid combined with normalized cross-correlation(NCC)function to detect blank matching positioning,and then using the fast Fourier transform in frequency domain processing images from zero defect background image.The defect difference image was obtained by background subtraction method and then the defect was extracted by threshold segmentation.Experiments show that the accuracy of defect detection is as high as 96.7%,and the detection time of each image is less than 0.054 s,which meets the requirements of industrial production speed.

关 键 词：缺陷检测 机器视觉 模板匹配 背景差影法 

分 类 号：TH16[机械工程—机械制造及自动化] TG501[金属学及工艺—金属切削加工及机床]