Abstract: Because of the slow speed of tracking and registration on scenarios of augmented reality， an algorithm was proposed based on the improved random fern on tracking and registration of augmented reality. The proposed algorithm has the offline training section and the online tracking section. A classifier based on the embedded fern was proposed. In the developed classifier， a supervised dimensionality reduction method and all possible information were used. What’s more， it was also used for the feature matching， and then camera pose was computed and virtual objects were rendered and registered. Experimental results show that the proposed algorithm is superior to other algorithms on the average classification accuracy. The average processing time of each frame is about 34.22ms， which can almost meet the real-time.

Key words: random fern, dimensionality reduction, classifier, tracking, augmented reality