In this paper, a 2-point algorithm is proposed to estimate the relative pose as well as the absolute scale between two vehicle-mounted cameras efficiently. The system model is deduced by combining a two-view geometric model and the planar motion constraint to reduce the degrees of freedom. 2-point correspondences are utilized to calculate the rotation information independently from the translation information, which indicates that the proposed algorithm can deal with pure rotation scenes. Besides, provided that the camera’s configuration satisfies a certain condition, the absolute scale can be recovered. An approximation algorithm is developed and combined with the random sample and consensus scheme to deal with the uneven ground surfaces in practice. As only 2-point correspondences are required, less iterations are demanded in the estimating procedure compared with many other existing related algorithms. Both simulation and experiments are implemented to evaluate the proposed algorithm, in which the synthetic data, virtual robot experimentation platform, KITTI Vision Benchmark, and SUMMIT-XL platform are acquired. According to the results, the proposed algorithm performs better than many related algorithms including the well-known 5-point algorithm in many cases, especially when the camera’s trajectory contains sharp corners.

中文翻译：

---

一种有效的平面运动下车载相机相对位姿恢复方法

在本文中，提出了一种2点算法来有效地估计两个车载摄像头之间的相对位姿和绝对尺度。该系统模型是通过结合两视图几何模型和平面运动约束来减少自由度来推导出来的。利用2点对应独立于平移信息计算旋转信息，这表明该算法可以处理纯旋转场景。此外，只要相机的配置满足一定条件，就可以恢复绝对比例。开发了一种近似算法，并结合随机样本和共识方案在实践中处理不平坦的地面。由于只需要两点对应，与许多其他现有相关算法相比，估计过程中需要的迭代次数更少。通过仿真和实验来评估所提出的算法，其中获得了合成数据、虚拟机器人实验平台、KITTI Vision Benchmark 和 SUMMIT-XL 平台。结果表明，该算法在很多情况下都优于包括著名的5点算法在内的许多相关算法，尤其是在相机轨迹包含尖角的情况下。