一种基于视觉测量的大型自由曲面拼接方法

针对不规则复杂大尺寸三维面型测量的问题，提出一种基于视觉测量技术的大型自由曲面拼接方法。将大型曲面分为若干子块，采用双摄像机面结构光方式获取局部子区域密集点云，在曲面上布置若干标志点，其坐标采取近景摄影测量方法获得，对这些标志点进行局部子块坐标系到全局坐标系的转换，得到自由曲面面型。避免了一般标志点拼接方法的累积误差，解决了非编码标志点匹配及匹配奇异问题。实验表明，在1065mm×1000mm×255mm范围内，用19块子区域进行拼接，平均拼接误差为0．045mm。

Registration method for large free-form profile based on vision measurement

A registration method for large free-form profile is proposed. The curved surface is divided into several sub-regions and compact 3-D points called point clouds in every sub-region are separately measured by structured light method. Some global marked points in the curved surface are obtained using close-range photogrammetry and all cloud points in different local coordinations are converted at the global coordination. This method can avoid cumulated errors becauseof using global marked points. The matching of marked points is settled based on the invariability of spatial character and matching ambiguity is eliminated. An object about 1065mm × 1000mm × 255mm with 19 sub-regions is measured and average registration deviation is 0. 045mm.