关键词: 本质矩阵分解, 相对定向, 多视图几何

Abstract:  For purpose of acquiring the pose and position information of two cameras from essential matrix decomposition efficiently, a novel analytical essential matrix decomposition method is proposed. Firstly, by taking essential matrix’s rank 2 property into account, the position information of the two cameras corresponding to two views respectively is acquired. Secondly, the pose information is obtained through solving rotation matrix equations. Finally, in order to distinguish the unique solution that satisfies the reference-point visibility constraint quickly, the 3D point’s imaging depths are obtained directly. The experimental results show that, the proposed method not only avoids the time consuming singular value decomposition operation, but also acquires four analytical solutions that have intuitive and explicit geometric meanings, meanwhile because of the avoidance of the scene structure’s 3D reconstruction, the process of unique solution determination is simplified greatly.

Key words: relative orientation, essential matrix decomposition, multiple view geometry