Research On Tracking And Registration Algorithm In Augmented Reality

An AR (Augmented Reality) system supplements the real world with virtual objects that appear to coexist in the same space as the real world. Unlike VR, users are not immersed in an artificial environment, but in the real world with objects or information superimposed. A user can see the real world with the superimposed virtual objects through displays, such as handheld monitors and head mounted displays. With a wide range of applications, AR technique is receiving more and more researchers'focus. This thesis focuses on the research of outdoor application, tracking and registration in AR system. The main research work of the paper is as follows:1. The research status, development and application of augmented reality are introduced firstly, and then the key technology of augmented reality is analyzed deeply, such as: tracking and registration in the process of augmented reality system.2. Firstly, some computer image processing technologies in the development of augmented reality are introduced. These technologies include image segmentation, edge extracting and so on; Secondly, we introduce the template matching technology based on square marker and feature matching technology based on feature points; Finally, the perspective of 3D to 2D in computer vision and graphics related technologies are analyzed deeply.3. Several commonly AR development kits are analyzed. In the process of recognition, ARToolKit adopts fixed threshold segmentation method that is not in favor of complex environment. In this thesis, we use an adaptive image segmentation algorithm and design a Multi-marker Augmented Reality (MAR) system. Based on the above work, we design and implement an outdoor navigation system according to user's interested images. The experiments indicate that our system has the advantage of certain anti-noise ability; therefore it can be applied into complex environment successfully. 4. A robust natural feature tracking and registration method is proposed on the above research. This thesis adopts SURF combined with RANSAC to get the homography matrix and employ KLT algorithm to track the natural objects. The experiment result proved it is a robust tracking and registration method.