Robust Points Tracking Method Using Euclidean Reconstruction for Augmented Reality

Natural feature tracking is a very important research topic in computer vision field and has been used widely in Augmented Reality (AR). This paper gives a robust points tracking or transferring method based on the Euclidean reconstruction technique for AR systems. The points to be tracked include the lost natural features, and any points that are specified by the users. The proposed method distinguishes itself in following ways: Firstly, it is stable as it remains effective even when the camera is moved rapidly. Secondly, the proposed method is robust because it can operate normally as long as at least four pairs of reference point correspondences can be found during the augmentation process. Thirdly, we propose an augmented optical flow method by which the registration, annotation and video augmentation can still work even under the circumstances of large changes in illumination and viewpoint during the entire process. Several experiments have been conducted to validate the usability of the proposed approach.