基于机器视觉的凸轮轴表面缺陷检测实验研究  

Experimental research on camshaft surface defectdetection based on machine vision

作　　者：吴兆中 张新娜 王栋 赵锦国 卢唯辰 WU Zhaozhong;ZHANG Xinna;WANG Dong;ZHAO Jinguo;LU Weichen(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China;College of Modern Science and Technology,China Jiliang University,Jinhua 322002,China;Engineering Training Center,China Jiliang University,Hangzhou 310018,China;Zhejiang Gaohe Precision Machinery Co.,Ltd.,Jinhua 321016,China)

机构地区：[1]中国计量大学机电工程学院,浙江杭州310018 [2]中国计量大学现代科技学院,浙江金华322002 [3]中国计量大学工程训练中心,浙江杭州310018 [4]浙江高和精密机械有限公司,浙江金华321016

出　　处：《机电工程》2025年第4期697-705,797,共10页Journal of Mechanical & Electrical Engineering

基　　金：浙江省重点研发计划项目(2019C01128);金华市科技计划项目(2020-1-030)。

摘　　要：目前在凸轮轴表面缺陷人工目视检测方面存在精度差、效率低的问题,为此,对凸轮轴表面缺陷检测技术进行了研究,搭建了基于机器视觉的的凸轮轴表面缺陷检测系统,提出了一种基于凸轮轮廓曲线的图像分段采集策略,针对凸轮和齿轮分别设计了缺陷检测算法,对检测准确率及系统运行稳定性进行了实验研究。首先,结合凸轮轮廓曲线确定了一种图像分段采集策略,改善了因凸轮厚度较大和形状不规则造成图像部分区域无效的问题,获得了图像采集的最佳角度与次数;然后,采用形态学梯度,强化了凸轮亮度不均图像缺陷特征,结合Canny算子进行了缺陷检测,提出了一种基于像素补充的凸轮边界缺陷识别算法;最后,设计了一种基于齿顶尺寸关系和仿射变换的齿轮齿顶提取方法,通过形态学的开操作弱化了齿顶内部纹理,通过固定阈值分割了齿轮齿顶表面缺陷;搭建了检测系统硬件平台,对凸轮轴表面缺陷检测进行了实验及结果分析。研究结果表明:系统运行流畅,各工位检测准确率最低达98.11%,平均检测时间为11.09 s,缺陷分类准确率为96%。该凸轮轴表面缺陷检测系统满足了凸轮轴产线在线检测的需求,有效提升了其检测效率与准确率。At present,there are problems of poor accuracy and low efficiency in manual visual inspection of camshaft surface defects.Therefore,the detection technology of camshaft surface defects was studied,a camshaft surface defect detection system based on machine vision was established,an image segmentation acquisition strategy based on cam contour curve was proposed,defect detection algorithms were designed for cam and gear respectively,and the detection accuracy and system operation stability were studied experimentally.Firstly,in combination with the cam contour curve,an image segmentation acquisition strategy was determined,ameliorating the problem of ineffective image regions caused by the considerable thickness and irregular shape of the camshaft,and attaining the optimal angle and frequency of image acquisition.Then,the morphological gradient was employed to intensify the defect features of the uneven luminance image of the cam,and the Canny operator was utilized for defect detection.A cam boundary defect recognition algorithm based on pixel supplementation was proposed.Finally,a gear tooth top extraction method based on the size relationship of the tooth top and affine transformation was designed.The internal texture of the tooth top was weakened through morphological opening operation,and the surface defects of the gear tooth top were segmented by fixed threshold.The hardware of the detection system was established to conduct experimental analysis of camshaft surface defect detection.The experimental results indicate that the system operates smoothly,with the lowest detection accuracy at each station reaching 98.11%,an average detection time of 11.09 seconds,and a defect classification accuracy of 96%.This system fulfills the requirements for online detection in the camshaft production line and effectively enhances detection efficiency and accuracy.

关 键 词：检测精度 凸轮轴轮廓曲线 齿轮齿顶缺陷检测算法 图像分段采集策略 特征强化 图像分割 

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]