基于机器视觉的连接器缺陷自动检测  

Automatic Detection of Connector Defects Based on Computer Vision

作　　者：陈永清[1] 张跃飞[2] 张维刚 程晓民[1] 陈廉清[1] 

机构地区：[1]宁波工程学院 [2]浙江工贸职业技术学院 [3]杭州威仕达机电科技有限公司

出　　处：《工具技术》2015年第6期78-81,共4页Tool Engineering

基　　金：国家自然科学基金(51275251);浙江省自然科学基金(Y14E050021)

摘　　要：针对人工检测连接器效率低、易漏检,严重影响产量与质量,提出一种基于机器视觉的非接触式自动检测方法。连接器的缺陷存在于本体的内外表面,选择合适的照明方案使缺陷都能被采集到;缺陷特征类型多、差异大,利用卷积滤波处理得到高质量的图像,缺陷特征与背景区别明显;最后,应用Otsu阈值分割处理,采用边缘直线建立坐标系的方法自动定位检测有效区域,消除了产品的位置误差;统计目标像素数,检测目标。实际检测表明,该系统检测结果稳定,正确率高,速度快,可满足连接器生产自动检测的实际需求。A non-contact automatic detection approach using computer vision is proposed to address the problems like low productivity, unreliable quality, inefficiency, and misdetection associated with manual detection. A lighting solution is employed to collect the defects on the inside and outside surfaces of the body. Concerning the diversity of defect features, convolution amplification is adopted to obtain high-quality image which clarifies the differences between defect and back- ground. Finally, with segmentation processing of Otsu threshold and the coordinate system from edge line, effective area is automatically detected and position error is removed, which allows count the pixels and detect the goal. Experimental detection shows that the proposed system has these characteristics of reliable performance, high correct rate, and high working speed, and can meet the requirements of automatic detection for connectors on the shop floors.

关 键 词：连接器 图像分割 滤波 缺陷检测 

分 类 号：TG115[金属学及工艺—物理冶金] TP391[金属学及工艺—金属学]