从二维图像直线光流场求解三维刚体旋转运动参数

Reconstructing 3D rotation motion parameters from 2D image straight-line optical flow

本文将图像直线的三个参数对时间的导数定义成直线光流场,找出了在透视投影模型下运动刚体上的空间直线与其投影的图像直线之间的关系,提出了一种利用单目图像序列中两幅连续图像的三对直线光流场,通过解线性方程组得到刚体旋转运动的算法,同时还可以得到摄像机的一个内参数焦距。由于是解线性方程组,无需迭代和给出迭代初值且所需要的直线数目少,所以该算法简单,运算速度较快,容易实现。

This paper defines the derivative to tlie three parameters of 2D llne which is the perspective projection of 3D line in motioning object as the straight - line optical flow. A new simple algorithm of reconstructing 3D rotation motion from 2D image straight - line optical flow is presented. The algofitlun is to only use three couple of straight - line optical flow in two consecutive image frames of image sequenece to get the 3D rotation motion parameters by solving the linear equations, at the same time, the focus of camera can be obtained. It needs no iteration and iterative initial value and requires less 2D straight - line optical flow, therefore, the algorithm is easier to be realized.