由双目图象流无对应地估计3－D运动参数

COMPUTING GENERAL 3-D MOTION OF OBJECTS WITHOUT CORRESPONENCE FROM BINOCULAR IMAGE FLOWS

本文设计用并行立体摄象机产生的图象流作为输入，通过解系数由图象速度矩和两图象内坐标构成的线性系统等式，不需左、石两图象流场间特征点与特征点的对应就能复原物体的３－Ｄ运动参数．经用综合图象流场测验已经证明我们算法的可行性以及具有功能强、速度快的特点．

This paper,uses the binocular image flows from parallel stereo cameras as input data for computing general 3-D motion.A linear system of equations,whose coefficients consist of the momonts of image velocities and coordinates in the two images,does not require point-to-point correspondence between the left and right flow fields.Experimental results with synthetic and real flow fields have demonstrated the feasibility of the scheme and some of its properties.