A model registration system capable of tracking an object, the model of which is known, in an image sequence is presented. It integrates tracking, pose determination and updating of the visible features. The heart of our system is the pose computation method, which handles various features (points, lines and free-form curves) in a very robust way. It consists in using robust estimators in a two-stage process: a robust residual is computed for each feature and a robust function of these residuals is minimized. Our method is therefore able to give a correct estimate of the pose even when tracking errors occur. This method is also used to update the set of model features along the sequence, by finding the 2D contours that are tracked when new 3D features become visible. The reliability of the system has been shown on a 300-image sequence of a bridge shot at dusk time, for an augmented reality project.