一种相机位姿鲁棒估计方法

针对位姿估计算法对鲁棒性和精度要求高的特点,提出了一种鲁棒的高精度位姿估计方法:该方法以3个样本点对为基础,首先假设各空间点到投影中心的距离近似相等,然后根据像点与空间点间的几何约束关系建立方程组并推导迭代公式,进而迭代求解各空间点到投影中心的距离;以此为基础,提出了一种最小二乘绝对定向方法来获取相机的初始位姿。最后根据空间点与像点之间的共线约束进行迭代优化,得到最终的相机位姿。测试表明,该方法在满足假设的条件下能很好地收敛,并达到较高的解算精度。

A Robust Method for Camera Pose Estimation

Against the requirement of robustness and accuracy of pose estimation ,a robust and accurate pose estimation method is proposed .The method is based on 3 sample point pairs .First ,the distance of every spatial point to the projection center is assumed to be nearly the same .Then the e-quation group ,from which the iteration formula is derived ,is constructed according to the geometry constraint relationship between the image points and the spatial points .Subsequently ,the distance between the spatial points and the projection center is solved iteratively .Based on the above result ,a least square absolute orientation method is proposed to solve the initial pose of the camera .Finally , the initial camera pose is optimized iteratively based on the colinearity constraint between the spatial points and the image points .Experiment results illustrate that ,in the assumption condition ,the pro-posed method can converge well and achieve high accuracy .