An approach to improve the robustness of estimating 3-D motion parameters from monocular and binocular image sequences

On the basis of the Cayley transform, this paper proposes a linear algorithm for determining 3-D motion parameters of a rigid body in monocular and binocular cases. Since the computation for rotation parameters is simplified from nine variables to three variables in the algorithm, the complexity of the computation is reduced so that the reliability and the stability are improved. Simulation experiments show that the relative error of the rotation matrix is up to 5%, while the relative error of the intermediate parameters is about 50%, caused by the truncation error of image data in the monocular case