Bilkent University
Department of Computer Engineering


Model-based Camera Tracking for Augmented Reality


Aytek Aman
PhD Student
Computer Engineering Department
Bilkent University

Augmented Reality (AR) is the enhancement of real scenes with virtual entities. To provide natural AR experiences, extrinsic camera parameters must be calculated in an accurate, robust and efficient way so that virtual entities can be overlaid onto the real environments correctly. Vision-based tracking serve as the main method for estimating the camera pose. For rich-textured environments, keypoint-based methods work well and heavily used. Edgebased tracking, on the other hand, is more preferable when the environment is rich in geometry but has little or no visible texture. Pose estimation for edge-based tracking systems generally depends on control points assigned on the model edges. For accurate tracking, visibility of control points must be determined correctly, which is computationally expensive. We propose a method to reduce the computational cost of edge-based tracking by preprocessing the visibility of the control points. For this purpose, we use persistent control points that are generated in the world space during the preprocessing step. We exploit tree-like structure of the control points to reduce memory overhead due to preprocessed visibility. Additionally, we use an adaptive projection algorithm for persistent control points to provide a more uniform control point distribution in the screen space. We test our camera tracker using a public data set. The preprocessed visibility information enables fast determination of control point visibility while preserving the accuracy of the tracker.


DATE: 22 February, 2016, Monday @ 15:40