机器视觉在网片缺陷检测与分类中的应用  

Application of Machine Vision in Detection and Classification of Mesh Defects

作　　者：顾寄南[1] 唐良颖 许悦 唐仕喜 GU Ji-nan;TANG Liang-ying;XU Yue;TANG Shi-xi(Mechanical Information Research Center of Jiangsu University,Jiangsu Zhenjiang 212013,China)

机构地区：[1]江苏大学制造业信息化研究中心

出　　处：《机械设计与制造》2019年第A01期47-49,53,共4页Machinery Design & Manufacture

基　　金：智能化铸件后处理成套设备的研发与产业化(BA2015026)

摘　　要：针对网片缺陷传统人工检测方法误检率高、劳动强度大等问题,应用机器视觉技术,提出了一种网片缺陷在线检测及分类方法。首先通过工业相机获取网片图像,应用中值滤波和图像二值化方法实现对网片图像的预处理。通过分析缺陷特征,提出了基于特征点的网片缺陷检测方法,在检测出缺陷的同时能对网片三种缺陷类型进行预分类。根据网片缺陷类型的不同,通过计算缺陷区域的灰度共生矩阵并提取4个特征参数,运用BP神经网络对网片缺陷进行分类。实验表明,使用本方法分类网片缺陷类型能满足工业要求。In view of the problems of high missing detection rate,labor intensiveness of traditional manual inspection methods of mesh defects,a method of on-line detection and classification of mesh defects is proposedby using machine vision technology.Firstly,the mesh image is obtained by industrial camera,and the preprocessing of mesh images is achieved by applying methods of median filtering and image binarization.By analyzing the defects types of the mesh,a mesh defects detection method based on feature points is proposed.The three kinds of defects types can be pre-classified while detecting the defects.According to the types of mesh defects,four characteristic parameters were extracted by calculating gray-level co-occurrence matrix of the defect area,the BP neural network was used to classify the mesh defects.Experiments show that using this method to classify mesh defects can meets industrial requirements.

关 键 词：机器视觉 网片 BP神经网络 缺陷分类 灰度共生矩阵 

分 类 号：TH16[机械工程—机械制造及自动化] S971.4[农业科学—捕捞与储运]