基于单目视觉的零件表面网格节点坐标测量

Extraction of Surface Grid Node Based on Monocular Vision

摘要: 本文提出一种新的基于几何运算的网格节点提取方法，首先对获得的图像进行二值化及均值滤波等处理，然后运用Halcon强大的算子函数库，获得网格线的亚像素精度，然后进行网格线的合并，像素点的提取，网格交点的提取等，用此种方法获得的零件表面网格节点图像坐标受灰度值的影响会比较小。最后对不同的实际零件进行平移实验，验证了基于平移的方法的可行性，为研究基于一般运动的零件的三维坐标测量提供了理论基础。

Abstract: In this paper, a new mesh node extraction method based on geometric operation is proposed. Firstly, the obtained images are binary and mean filtered, and then the sub-pixel accuracy of the mesh line is obtained by using Halcon powerful operator function library. Then the mesh line is merged, pixel point is extracted and mesh intersection is extracted. Finally, the translation ex-periments of different practical parts are carried out to verify the feasibility of the translation-based method, which provides a theoretical basis for the study of three-dimensional coordinate measurement of parts based on general motion.