A markerless 3D tracking approach for augmented reality applications

Nowadays, augmented reality (AR) applications are still dominated by marker-based solutions to perform pose estimation. The main reason for this is that existing marker-less trackers are still not robust enough to serve the purpose of tracking the head of the user at all times. In this paper, we propose an improvement over existing camera trackers by using only depth information acquired by hand-held depth cameras. The main contribution of this paper is the way we combine a dense tracking approach with a model-based one. More specifically, besides pose estimation based on the alignment of consecutive point clouds, a second thread is running for the alignment of the acquired point cloud with a 3D model. To make the tracking both robust, accurate and fast, we propose a surface reconstruction step combined with a multi-resolution resampling of the 3D model. Our system was extensively tested on an online available dataset and the accuracy proved to be competitive with state-of-the-art visual SLAM methods.