Abstract:This paper presents a novel approach for depth and motion estimation,aiming at the model-based pose estimation problem.The estimation procedure consists of two steps.Firstly,it estimates the rotational matrix according to three point pairs constituted by three model points.Then,the translation vector and depths are computed from one of the three pairs.In order to enhance the precision of the estimated results,a simple and effective optimization algorithm is also proposed.The main advantages of this method are that it estimates the rotational matrix individually and can obtain good estimation results according to small number of model points.Experiments performed on synthetic data and real images demonstrate the effectiveness of this approach.
[1] Tagawa N,Inagaki A,Minagawa A. Parametric estimation of optical flow from two perspective views[J]. IEICE Transactions on Information and Systems,2001,E84-D(4): 485-494. [2] Park S K,Kweon I S. Robust and direct estimation of 3-D motion and scene depth from stereo image sequences [J]. Pattern Recognition,2001,34(9): 1713-1728. [3] Ho A,Pong T C. Cooperative fusion of stereo and motion [J]. Pattern Recognition,1996,29(1): 121-130. [4] Nuno G,Helder A. Estimation of 3D motion from stereo images - uncertainty analysis and experimental results [A]. Proceedings of the IEEE International Conference on Intelligent Robots and Systems[C]. 2002. vol.1. 7-12. [5] Oberkampf D,Dementhon D F,Davis L S. Iterative pose estimation using coplanar feature points [J]. Computer Vision and Image Understanding,1998,63(3): 495-511. [6] Or S H,Luk W S,Wong K H,et al. An effective iterative pose estimation algorithm [J]. Image and Vision Computing,1998,16(5): 353-362. [7] Wang F. A simple and analytical procedure for calibrating extrinsic camera parameters [J]. IEEE Transactions on Robotics and Automation,2004,20(1): 121-124. [8] Newman P M. On the structure and solution of the simultaneous localisation and map building problem [D]. Australian: The University of Sydney,1999.