Markerless Augmented Reality System Based on Face Tracking

Abstract

Traditional vision-based augmented reality systems rely on a marker. The marker has to be encoded in a standard way and located in the sensing range. In this paper, instead of a card board marker, we design an AR system using human face as a distinctive pattern for tracking. We first apply our tracking algorithm to detect and track the feature points in the face area, and frame-by-frame reconstruct the camera pose relative to the face by analyze the feature points, finally draw a 3d mustache on the face upper lip area. Our tracking algorithm is based on the original ASM fitting method but we combine it with optical flow tracking. We evaluate our approach with regard to speed and accuracy

the result proves that the performance of our proposed system is very robust.