Augmented Reality(AR)is the merging of synthetic sensory information into a user's perception of a real environment. Based on camera models and calibration technology, this paper presents thorough discussions on the key technology of AR system based on vision. Our object is improving the accuracy of tracking registration problem.Firstly, according to the research of AR registration and analysis of vision servo, we conclude that we can make use of the theory of vision servo to solve this key problem and applied to camera calibration. We research a calibration way based on virtual vision servo. Compared with other calibration methods, this method is more flexible and reliable.Secondly, based on accurate camera calibration, we realize a new vision-based tracking algorithm based on natural feature point. Robust tracking is obtained by merging statistics into control law. Comparisons with other typical algorithms are also given.Thirdly, we research model-based and moving matching-based makerless tracking system. A new tracking system based planar structure in the scene is proposed. This system can easily be used for Augmented Reality outdoor environment applications. |