A new homography decomposition algorithm and its application in visual servoing control

In this paper, a new singular value decomposition based numerical algorithm is proposed to solve the homography decomposition problem. Compared with traditional homography decomposition methods, the algorithm gives not only the relationships among eight mathematical solutions, but also the relationships among the reconstructed structure of the scene. And the set of solutions of the homography decomposition problem can be reduced from eight mathematical solutions to the only two verifying a physical constraint. Finally during the off-line learning stage, the unknown desired vector dn between desired camera frame and 3D plane can be estimated, the unique solution of the homography decomposition problem can be distinguished. At the end of this paper, a 2-1/2-D vision-based recognition experiment of the QR Code landmark for mobile robot is designed using the unique solution of the proposed algorithm, the experimental results show that the homography decomposition algorithm is simple but efficient. © 2009 by Binary Information Press.