Position estimation of non deformable objects based on butterfly-topology parallel GA

Abstract

Motion estimation is a very important problem in dynamic scene analysis. Although it is easier to estimate motion parameters from 3D data than from 2D images, it is not trivial since the 3D data we have are almost always corrupted by noise. We address the problem of computing the three-dimensional motion of objects. This paper proposes a robust approach to position estimation of moving objects by exploiting the only available geometric constraint, namely, the epipolar constraint. The extrinsic parameters of the camera, and the motion of the stereo rig is unknown. If we make an exhaustive search for the epipolar geometry, the complexity is prohibitively high. The idea underlying our approach is to use a parallel fine-grain GA as an optimizer. Since the constraint on the rotation matrix is not fully exploited in the analytical method, nonlinear minimization can be used to improve the result. In our experiments, computer simulated data is used to validate the proposed algorithm, and very promising results are obtained.