Stereo Camera Motion Estimation Based on Euclidean Reconstruction

Motion estimation for an unknown parameters stereo camera rig is an important topic in the study of computer vision. This paper have proposed a method based on Euclidean reconstruction. Firstly, some feature points are extracted and matched. Then, they are reconstructed into the points in the 3D coordinates up to a scale, which is the length of the baseline. The unknown focal length is the key parameter to be solved in the process of reconstruction. It is estimated using the constraints of the reconstructed structures in Euclidean space. After that, the reconstructed points are determined. The problem is then transformed to solve the rigid transformation between two sets of 3D points. A non-linear optimization is adopted at last. Experimental results show that our method is simple, effective and accurate.