Abstract: In the mobile augmented reality, because the low performance of mobile terminals leads to that the capability of feature extraction matching and tracking recognition is limited, this paper uses the FAST-SURF algorithm to detect and describe the plane marker, to save the feature points into the structure of KD-Tree to accelerate the matching speed with the offline collected feature points, through the KLT tracking algorithm to track the key feature frame to determine the position and orientation of the camera, and to superimpose the three-dimensional virtual items onto the real scene. The transplantation experiments of the algorithm on the Android platform show that, in the case of different angles, certain environmental changes and that the reference image is partly occluded, the system is of good display tracking performance.

Key words: mobile augmented reality, FAST-SURF, KD-Tree, Android