Abstract: In order to achieve real-time positioning of the workpiece in the industrial processes, this paper  proposes a real-time location and matching method for the workpiece with circular geometric primitives. In the offline process, it uses the polygonal approximation method to approximate the edge contour image, and further divides it into segments as line primitives and arc primitives. Templates are created. And it determines the position and orientation of the workpiece. In the online process, the longest circular arc is examined first; with it the template is transformed to the image. The primitives can be matched to the image based on the distances of the image primitives and the transformed template primitives. Tests show that this method for circular geometry primitives of the workpiece, in the case of occlusion, can quickly and accurately perform real-time matching location.

Key words: machine vision, geometric primitives, polygonal approximation, positioning and matching

摘要： 为实现工业过程中对工件的实时定位，提出了一种基于圆弧基元的工件实时定位与匹配方法。离线过程中，将边缘轮廓多边形近似，进一步分割成线段基元和圆弧基元，制作模板，确定模板中工件的位姿。在线过程中，以最长圆弧基元得到潜在匹配位置并变换模板，通过计算测试图像中几何基元与变换后的模板中对应的几何基元之间的距离来实现匹配定位。试验表明，该方法对于具有圆弧基元的几何形状的工件，能快速、准确地完成实时匹配定位。

关键词: 机器视觉, 几何基元, 多边形近似, 匹配定位