一种平面轮廓分割与基元识别方法研究  

Research on the method of planar contour segmentation and recognition of the geometric primitives

作　　者：李欣言 杜文华[1] 王俊元[1] 段能全[1] 关波[1] 汤恒[1] 李博[1] 

机构地区：[1]中北大学机械与动力工程学院,山西太原030051

出　　处：《制造技术与机床》2016年第1期85-88,共4页Manufacturing Technology & Machine Tool

基　　金：山西省研究生优秀创新资助项目(20133095)

摘　　要：为提高视觉测量系统的检测精度与检测速度,提出了一种平面轮廓分割与识别算法。首先对经过多边形逼近算法进行轮廓分割获得的分割点位置按一定规则进行修正;然后对分割点间的线段类型加以识别,对过度分割的圆弧段进行融合处理;最后用加权最小二乘拟合算法抑制了大的离群值的影响,将轮廓拟合成相对应的几何基元。对提出的方法分别进行检测精度与检测速度对比实验和基元识别能力测试实验。实验结果表明,提出的方法基元识别准确、检测速度快、通用性好。In order to improve the detection accuracy and speed of the vision measurement system,a method of planar contour segmentation and recognition is proposed. First,the position of the points of division obtained by the polygon approximation algorithm is modified by a certain rule. Then the types of line segments between the points are identified,and the over segmented arc is combined. In the end,the affect of large outliers is limited by the weighted least squares fitting algorithm. And the contour is synthesized into the corresponding geometric elements. Experimental demonstration of the proposed method is carried out by testing accuracy and speed detection contrast experiments and identification ability of basic element experiments. Experimental results show that the proposed method has the characteristics of high accuracy,fast detection speed and good generality.

关 键 词：机器视觉 轮廓分割 基元识别 多边形逼近 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]