A new approach for 3D motion estimation of a planar object

作者：Wang Chundong^*; Ning Shukun; Wen Xianbin; Zhang Hua

来源：Journal of Computational Information Systems, 2010, 6(5): 1517-1524.

• 2-piece
• 3D motion estimation
• 3D motion parameter estimation
• 3D-motion parameters
• 8-point-algorithm
• A-plane
• Annealing algorithm
• Estimation results
• Existing method
• Feature point
• Homography matrices
• Least squares methods
• Linear estimation
• New approaches
• Nonlinear approach
• Planar object

This paper proposes a new nonlinear approach for 3D motion estimation of a planar object. Based on the premise that the object is a plane or a kind of similar plane, we use LSM (Least Squares Method) to estimate the 3D motion parameters. Because the derived equations could not be solved deterministically, the steepest descent method and the annealing algorithm are used to find the solutions. Comparing to the 8-point-algorithm, one of the popular existing methods, our method needs much less corresponding feature points (at least 4 pairs are enough); comparing to the linear estimation methods, such as the homography matrix decomposition, our method needs much less information (only the coordinates of the corresponding feature points on 2 pieces of images) and the estimation result has no ambiguity problem. In the experiment, the result shows a high and stable accuracy.

• 出版日期2010
• 单位天津大学