Christophe Meilhac and Chahab Nastar,
A Robust and Precise Approach for Model-Based 3D/2D
Registration and Tracking,
In The 10th Scandinavian Conference on Image Analysis,
Lappeenranta, Finland, pp. 635-642, June 1997.
Christophe Meilhac and Chahab Nastar,
A Robust and Precise Approach for Model-Based 3D/2D
Registration and Tracking,
In The 10th Scandinavian Conference on Image Analysis,
Lappeenranta, Finland, pp. 635-642, June 1997.We describe a precise method for registering and tracking a 3D geometrical model in a 2D video sequence. The algorithm is based on an ``Iterative Closest Point'' technique. Each iteration is composed of two steps: computation of camera parameters, and 3D/2D vertex matching. This last step is performed by polygon fitting in an edge image. To account for false matches, we use a robust M-estimation both for camera parameter estimation and 2D feature extraction. Experimental results show that precise registration can be obtained even with very noisy outdoor images and incomplete data.