一种三维运动参数的鲁棒估计方法

A NEW ROBUST ALGORITHM FOR ESTIMATING 3-D MOTION PARAMETERS

本文提出了一种估计三维运动参数的线性算法,利用两帧图象之间点匹配的信息,算法给出了鲁棒性较好的解。该算法的关键之处是引入必要的先验知识,使得先验知识直接参与数值计算,以解决所谓不适定问题,实验结果表明:本文的算法在多方面优于现存的算法。

A new linear algorithm is presented for estimating 3-D motion parameters of rigid objects. Based on the print correspondence of two view sequential images, the more robust parameters (siainly rotation parameters) can be obtained. The key to the algorithm is to introduce some necessary priori and combine them with a stable numerical algorithm in order to solve some ill-posed problems. It is shown that the algorithm has advantages over the other existing ones.