一种快速多视拼接方法研究

A Quick Multi-view Registration Method Research

针对3D扫描测量领域自动化水平不高,测量不稳定,测量误差较大的现状,在双目结构光测量系统基于空间点相对距离不变性的多视拼接中,提出了一种基于三角约束和全局优化的多视拼接方法。该方法包括:匹配标记点时对由点距约束求得的可能匹配点加入三角约束,去除误匹配点;初次拼接后,根据求得的重心用最小二乘法进行全局优化,得到优化后的拼接矩。通过VC++,Open CV等工具的算法实现表明:该方法能够较好地抑制累积误差,拼接方法稳定、快速,能实现自动化拼接。

The measurement technology of 3D scanning in automation level is low, its measurement process is not stable, and themeasurement error is large. After the principle of binocular structured light measurement was analyzed, match the marked point tomulti-view registration based on the relative distance of any two marked point in space is unchanged. A multi-view registration isproposed using triangle constraint and global optimization. The method includes: remove the false matching points in possiblematching points by joining triple constraint; after the first registration, use the center of marking points to global optimizationthough least squares principle method. The algorithm is written by VC++ and Open CV, and the algorithm is carried out by experiments test. The measurement results show that this method has advantage of restrain the accumulation error, keep up the processstable, fast and automatic.