Planar Motion Estimation for Multi-camera System

Xinlei Qi, Yaqing Ding, Jin Xie, Jian Yang

Abstract: In this paper, we propose efficient solutions to relative pose estimation with a multi-camera system. We focus on the case where the system navigates under planar motion, and propose two new algorithms: the non-minimal linear 6-point algorithm and the minimal 3-point algorithm. For the 6-point algorithm, we use a simple and easy-to-implement way to avoid the SVD (singular value decomposition) induced degenerate configuration, which happens in the multi-camera system based relative pose estimation. The minimal 3-point algorithm results in a system of polynomials with respect to three unknowns, and we show that it can be converted to solve a univariate polynomial in degree 4. The proposed algorithms are compared with the state-of-the-art methods on both synthetic data and public real-world images. Experimental results show very promising performance in terms of accuracy, robustness and efficiency.