基于视觉原理的链轮缺陷检测方法研究  

Study of Sprocket Defect Detection Method Based on Principle of Visual

作　　者：姜都 刘永 王宸 李伟杰 陈沛森 王红霞 JIANG Du;LIU Yong;WANG Chen;LI Weijie;CHEN Peisen;WANG Hongxia(School of Mechanical Engineering,Hubei University of Automotive Technology)

机构地区：[1]湖北汽车工业学院机械工程学院

出　　处：《仪表技术与传感器》2025年第2期34-39,共6页Instrument Technique and Sensor

基　　金：国家科技重大专项项目(2018ZX04027001);湖北省高等学校优秀中青年科技创新团队计划项目(T2022027)。

摘　　要：针对现有的链轮缺陷检测方法存在效率低、易受主观因素影响等问题,提出一种基于视觉原理的链轮缺陷检测方法。首先,采用掩膜法与连通域分析,分类检测出轮齿缺陷;其次,使用图像金字塔对待检图像进行采样;然后,提出一种角度偏差计算与搜索空间限制的改进匹配算法提升检测效率;最后,通过计算模板与待检图像之间的Hausdorff距离,检测出轮廓缺陷。实验结果表明:轮齿缺陷的检测准确率为98%,轮廓缺陷的检测准确率为95%,平均检测时间分别不超过2、9 s,可以满足自动检测的需求。In view of problems of the existing sprocket defect detection methods such as low efficiency,easily influenced by subjective factors,a kind of sprocket defect detection method based on the visual principle was put forward.Firstly,the mask method and connected component analysis were used to classify and detect the gear tooth defects.Then,the image pyramid was used to sample the image to be inspected.Secondly,an improved matching algorithm with angle deviation calculation and search space limitation was proposed to improve the detection efficiency.Finally,the contour defects were detected by calculating the Hausdorff distance between the template and the image to be detected.The experimental results show that the detection accuracy of gear tooth defects is 98%,the detection accuracy of contour defects is 95%,and the average detection time is not more than 2 s and 9 s respectively,which can meet the needs of automatic detection.

关 键 词：视觉原理 缺陷检测 金字塔分层算法 模板匹配算法 HAUSDORFF距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]