基于图像识别的机械齿轮加工缺陷识别方法研究  

Research on machining defect recognition method of mechanical gear based on image recognition

作　　者：陶文东[1] 朱亮亮[1] 付建军[1] TAO Wendong;ZHU Liangliang;FU Jianjun(Yangling Vocational&Technical College,Yangling Shanxi 712100,China)

机构地区：[1]杨凌职业技术学院,陕西杨凌712100

出　　处：《自动化与仪器仪表》2022年第9期62-65,共4页Automation & Instrumentation

基　　金：《2021年度陕西省职业技术教育学会课程思政项目》(SGKCSZ2020—562);杨凌职业技术学院2020年院内基金项目《基于CAD/CAPP智能化的零件加工工艺规程设计与应用》研究成果(ZK20—46)。

摘　　要：机械齿轮加工缺陷识别过程中,多以图像像素差异为依据,使识别结果AP值较低。为解决这一问题,提出基于图像识别的机械齿轮加工缺陷识别方法研究。针对机械齿轮加工过程,建立数字化测量方案。通过数字摄像头采集齿轮图像,并进行噪声过滤、边缘提取和图像分割。依托图像识别技术,引入核主成分分析原理,提取齿轮图像中包含的缺陷特征值,作为缺陷识别的依据。运用支持向量机方法,对特征值进行分类处理,得到齿轮加工缺陷识别结果。实验结果表明:所提识别方法与三种对比方法相比,机械齿轮加工缺陷识别结果的AP值分别提升了11%、13%、15%。In the process of mechanical gear machining defect identification, the difference of image pixel is usually taken as the basis, which results in low AP value. To solve this problem, an image recognition method based on mechanical gear machining defect recognition was proposed. A digital measurement scheme is established for the machining process of mechanical gear. The gear image is collected by digital camera, and the noise filtering, edge extraction and image segmentation are carried out. Based on the image recognition technology and the principle of kernel principal component analysis, the defect eigenvalues contained in the gear image are extracted as the basis of defect recognition. Using the support vector machine method, the eigenvalues are classified and processed, and the recognition results of gear machining defects are obtained. The experimental results show that compared with the three methods, the AP value of mechanical gear machining defect recognition results improved by 11%, 13% and 15%, respectively.

关 键 词：图像识别 机械齿轮加工 缺陷识别 CAD/CAM CAPP 机械零件加工工艺 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]