A registration method based on texture tracking using ARToolKit

In this paper, we propose a computer-vision based registration method for augmented reality based on template matching. Computer-vision tracking methods for augmented reality applications typically use a special fiducial markers such as squares or circles. Our new method uses a black square fiducial of ARToolKit to obtain the initial tracking condition, but does not use it in subsequent iterative tracking phases. Several natural feature points are extracted from the tracked object by offline image analysis. While tracking the object, some of these feature points are selected for template matching and the object pose and position are calculated. Even when the initial tracking square is no longer in view, our method produces robust tracking at real time frame rates.