基于混合优化的自适应加速稳健PnP算法

An Adaptive Accelerating Robust PnP Algorithm Based on Hybrid Optimization

根据目前摄像机位姿估计应用的实时性要求,针对RPnP算法在利用最小二乘误差求解时无法得到唯一解的问题,提出了一种改进自适应加速RPnP算法。在计算误差最小二乘时加入更多的限制条件,确定了输出解的唯一性;随后计算对应的摄像机外参数,代替每个极小值重投影误差的计算和比较过程,节省了大量位姿求解的时间;最后自适应地将原算法和改进后的算法相结合,使得输出结果最优化。实验证明,该方法可以大大降低算法的时间复杂度,并且运行时间受点数影响极小,可以较好地应用于实时性要求高的场景。

In view of the real-time requirement of the application of camera position estimation,and of the defect of the RPnP algorithm that it cannot get a unique solution when using the least square error,an adaptive accelerating RPnP algorithm is proposed.When calculating the error,more constraints are added to the least square,and the uniqueness of the output solution is guaranteed.Afterwards,the corresponding camera parameters are calculated,which are used to replace the calculation and comparison process of each minimum reprojection error,and thus can save a lot of time.Finally,the original algorithm and the improved algorithm are adaptively combined to optimize the output results.The experiment shows that this method can greatly reduce the time complexity of the algorithm,and the running time is hardly influenced by the number of points,which is more adaptable to the scene with high real-time requirement.