基于单应矩阵的相对位姿改进算法

摘要

The relative pose solution from homography matrix is vital for the visual navigation and visual servo applications. This paper proves new properties of the homography matrix and utilizes them to improve relative pose solution by the decomposition of homography matrix. Compared with previous methods, it only has a half of candidate solutions and enlarges the applicability range. Theory analysis, synthetic and real data tests show that the improved method saves 50%time in choosing the right relative pose from the candidate solutions and improves the total computational efficiency.%从单应矩阵恢复相对位姿在视觉导航、视觉伺服应用中具有重要价值,证明了单应矩阵的新性质,并利用该性质改进了基于单应矩阵分解的相对位姿估计算法.与已有算法相比,该算法的候选解个数减半,并扩大了适用范围.理论分析、合成数据和真实数据测试均表明,改进方法使从候选解中筛选出唯一解的运算时间减少了50%,提高了位姿估计的总体运算效率.