一种基于特征点稳定跟踪的三维注册方法

A 3D dimensional registration method based on feature point stable tracking

针对增强现实中三维注册易失效,稳定性差等问题,提出一种基于特征点稳定跟踪的三维注册方法。通过全局ORB特征点匹配得到目标特征点,采用光流法对目标特征点进行持续跟踪,并同时使用ORB特征对跟踪点进行校正。利用基于单应性矩阵的三维注册的特性恢复跟踪失效的特征点,最后利用相邻帧之间特征点的匹配求得单应性矩阵,进而求得注册矩阵。实验结果表明该方法能够基本满足三维注册实时性的要求,并且能够在复杂的注册环境中保持较好的鲁棒性。

Aiming at the problems of 3D registration in the augmented reality,such as failure and poor stability,this paper proposes a 3D dimensional registration method based on feature point stability tracking.Firstofall,the target feature points are obtained by global ORB feature point matching,and the target feature points are continuously tracked by the optical flow method,and the tracking points are corrected by using the ORB features at the same time. The characteristics of the 3D dimensional registration based on the homography matrix are used to recover the feature points of the tracking failure. Finally,the homography matrix is obtained by matching the feature points between adjacent frames,and then the registration matrix is obtained. The experimental results show that the method can meet the requirements of 3D dimensional registration real- time and maintain good robustness in complex registration environments.