Improvement of accuracy for 2D marker-based tracking using particle filter

This paper presents a method for improving accuracy of marker-based tracking using a 2D marker for Augmented Reality. We focus on that tracking becomes unstable when the view direction of the camera is almost perpendicular to a marker plane. Especially, tracking of Z axis which is perpendicular to the marker plane (X-Y) becomes unstable. For improving tracking accuracy in this case, we search rotation parameters that are the fittest to projected pattern based on the particle filter. By using particle filtering technique, then, our method can correctly estimate rotation parameters of the camera which are important to track the 3D coordinate system and improve the accuracy of the 3D coordinate system. This method can reduce jitters between frames, which is a big problem in AR. In the experiment, we demonstrate that our method can impivve the tracking accuracy of the 3D coordinate system compared with just using ARToolkit.