二值加权正交迭代相机位姿估计算法

A Binary Weighted Orthogonal Iterative Camera Pose Estimation Algorithm

在相机位姿估计算法的实际应用中,由于2D像点与3D参考点的错误匹配,参考点中往往含有异常值。针对现有算法抗异常值能力较弱的问题,提出了一种基于正交迭代的二值加权算法。该算法在正交迭代算法的基础上,引入包含两个阈值的加权系统来衡量参考点的可靠性。选取重投影误差的四分位数中的最大值作为阈值之一,另外引入一个与重投影误差和焦距相关的阈值来加速异常值比例较低时的收敛速度。最终选取两阈值的最大值对目标函数进行二值加权,降低了异常值对估计结果的影响。实验表明,该方法提升了正交迭代算法的抗异常值能力,具有较强的鲁棒性。

In the practical application of camera pose estimation algorithm,affected by the matching error,the reference points often contain outliers.In view of the weak ability of existing algorithms to resist the outliers,a binary weighted algorithm based on orthogonal iteration is proposed.Based on the orthogonal iterative algorithm,a weighted system with two thresholds is introduced to measure the reliability of the reference point.The maximum value of the quartile of the reprojection error is selected as one of the thresholds.In addition,a threshold related to the reprojection error and focal length is introduced to accelerate the convergence speed when the proportion of outliers is low.Finally,the maximum value of two thresholds is selected to weight the objective function,which reduces the influence of outliers on the estimation.The experimental results show that the method improves the anti-outlier ability of the orthogonal iterative algorithm and performs better in robustness.