一种高精度鲁棒的基于直线对应的位姿估计迭代算法

2D-3D特征对应位姿估计问题算法多基于点对应，而基于直线对应求解比基于点对应求解更具有优势。从欧氏空间这一新颖角度出发，提出了一种从直线对应求解位姿估计问题的迭代算法，算法思想是先迭代求解出最优的旋转矩阵，然后再得到平移向量。针对不同的直线组情形，给出了相应的迭代初始值计算方法。仿真实验数据表明，算法具有对初始值较不敏感、高精度、鲁棒性好等特点。

Accurate and robust iterative pose estimation from line correspondences

The methods for pose estimation problem from 2D to 3D feature correspondences are mainly based on point correspondences. However, it is advantageous to compute pose from line correspondences. From a novel point of view based on the Euclidean space, we presented an iterative algorithm for solving pose estimation problem by using line correspondences. The main idea is that, firstly the best rotation matrix is iteratively computed, and then the translation vector is obtained. For different groups of line correspondences, we presented different methods for getting an initial guess. Synthetic experimental data show the proposed algorithm is less sensitive to initialization, accurate and robust.