Abstract:

Binocular vision sensor could not obtain the 3D information of a surface of  the large-sized object by single measuring, due to the limitation of camera view field, a new measuring method was researched and finally achieved, that measured large-sized objects by using global control net. The experimental results demonstrate that this method possesses high-speed and high-precision, in addition it extends binocular vision sensor's measurement range indirectly. In another way, this method overcomes the shortcomings of low-precision effectively in the course of splicing the images in 3D shape measurement of large objects. This method is appropriate for large-scale measurement, it can obtain the relative location and the information of 3D dimensions for the characteristic points on the object rapidly, simply and accurately,
so that it can recovery 3D shape of objects.

Key words: large-sized 3D shape measurement, binocular vision theory, global control net, adjustment algorithmic, data fusion

摘要：

由于传感器成像视场范围的局限性,双目视觉传感器在一次成像中往往不能获得大尺寸物体的整体形貌测量结果,鉴于此,研究并实现了一种通过建立全局控制网来实现对大尺寸物体的三维形貌进行测量的方法。实验结果表明：基于全局控制网的测量方法速度快、精度高,间接地扩大了双目视觉传感器的测量范围,并且有效地克服了大尺寸三维形貌测量拼接精度低的缺点。该方法适用于大尺寸物体的三维形貌测量,能够快速、简单、精确地获得被测物体表面特征点的相对位置关系和三维尺寸信息,进而恢复被测物体的三维形貌。

关键词: 大尺寸三维形貌测量, 双目视觉原理, 全局控制网, 平差算法, 数据融合