基于支持向量回归的圆锥螺纹检测  

Conical thread detection based on support vector regression

作　　者：于忠党[1] 王龙山[1] 陈向伟[1] 

机构地区：[1]吉林大学机械科学与工程学院

出　　处：《辽宁工程技术大学学报（自然科学版）》2006年第3期432-435,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金资助项目(10572053);吉林大学创新基金资助项目(2004-1)

摘　　要：以支持向量回归为主要算法,讨论了圆锥螺纹各参数的图像检测方法。采用边缘保持滤波、二值变换等算法,对圆锥螺纹图像进行处理,获得牙形直线部分的像素表征,并以此构成训练集,进行支持向量回归,得到了螺纹牙形直线方程的亚像素表示,据此对锥螺纹的主要参数进行检测,大大降低了CCD的离散性和系统噪声对测量结果的影响。实验表明,本方法具有测量速度较快,测量精度较高的特点。The method of the conical thread image detection based on the support vector regression is presented. The pixel characterization of the thread form＇s straight portion can be gotten by processing the conical thread image with the algorithm of filter, binary transformation etc, which makes up of the training set, then, trains the Support Vector Machine for Regression and gets the straight line equation of the thread form which is the expression of the sub-pixel. The discreteness of the CCD and the system noise would be reduced greatly during the detection of the screw thread main parameters. The experiments show this method has some advantages of great detection speed and higher measure precision.

关 键 词：图像检测 CCD 圆锥螺纹 支持向量回归 

分 类 号：TH161[机械工程—机械制造及自动化]