基于机器视觉的注塑齿轮缺陷检测系统研究  

Injecting plastic gear defect detection system based on machine vision

作　　者：陈硕[1] 林洪毅 卞抱元 赵紫阳 唐旭晟[1] CHEN Shuo;LIN Hong-yi;BIAN Bao-yuan;ZHAO Zi-yang;TANG Xu-sheng(School of Mechanical Engineering and Automation,Fuzhou University,Fuzhou 350116,China)

机构地区：[1]福州大学机械工程及自动化学院,福建福州350116

出　　处：《机电工程》2022年第9期1243-1249,共7页Journal of Mechanical & Electrical Engineering

基　　金：国家科技部创新方法工作专项资助项目(2020IM010200)。

摘　　要：在对齿轮缺陷进行人工检测的过程中,存在检测效率低、人为因素影响大等诸多问题,为此,提出了一种基于机器视觉技术的水表齿轮缺陷检测方法。首先,提出了一种水表齿轮的视觉检测系统的总体方案,并据此构建了检测系统的三维模型;其次,采用中值滤波方式对采集到的图像进行了图像平滑处理,并运用Canny算子完成了对齿轮边缘轮廓的提取;然后,采用图像金字塔与形状匹配相结合的模板匹配算法,找到了齿轮边缘轮廓和模板的位置关系,利用放射变换算法将两者重心变换重合;最后,通过计算两者的Hausdorff距离,提取了模板和待检测齿轮外轮廓间的最大误差以判别次品;并通过搭建测试平台对实际生产的齿轮进行了测试实验,结合实验结果对齿轮缺陷检测系统进行了验证。研究结果表明:该齿轮缺陷检测系统能有效检测出齿轮存在的缺陷,且其误检率低于2%,漏检率低于1%,检测时间不超过40 ms,可满足齿轮缺陷的在线检测要求。Aiming at the problems of low efficiency and high influence of human factors in manual inspection of gear defects,a detection method of water meter gear-defect based on machine vision technology was proposed.Firstly,the overall scheme of the water meter gear visual inspection system was designed,and a three-dimensional model of gear visual inspection system was constructed.Secondly,the median filtering method was used for image smoothing processing of the collected images and the edge extraction was done by using the Canny operator.Then,the template matching algorithm combined image pyramid and shape matching was used to find the position relationship between the gear edge contour and the template,and both of their center of gravity could be overlapped by using affine transformation operator.Finally,the maximum error between the template and the outer profile of the gear to be detected was extracted by calculating the Hausdorff distance between the two to identify defective products.A testing platform was established to test the actual gears and verify the gear defect detection system.The experimental results show that the proposed method can effectively detect the defects in the bearing end face.The false detection rate is less than 2% and the undetected rate is less than 1%.The detection time is less than 40 ms.The proposed algorithm can meet the requirements of on-line inspection.

关 键 词：齿轮传动 中值滤波技术 CANNY算子 齿轮边缘轮廓提取 齿轮缺陷在线检测 模板匹配算法 HAUSDORFF距离 图像平滑处理 

分 类 号：TH132.41[机械工程—机械制造及自动化] TP391.4[自动化与计算机技术—计算机应用技术]