A distributed tracking system for indoor augmented reality applications

The indoor distributed tracking system proposed in this paper is a client-server architecture with three basis subsystems: the site information server, mobile units, and the network infrastructure. The tracking process is divided into four procedures: camera pose computing, 2D features tracking, 2D-3D correspondence and natural feature calibration. To achieve real-time behavior required for tracking applications, dynamic interpretation tree is used to organize data in a multimedia object database. The computations of 2D image registration for pose estimation are done on basis of the Fast Fourier Transform because of its robustness and fast calculation. To track camera pose and dynamically estimate the 3D positions of natural feature, we use an auto-calibration approach based on an iterative Extended Kalman Filter. The result shows that our system can successfully provide indoors mobile users with 3D positions from 2D image points and information of augmented reality.