抄録
We present a novel vision-based pose detection method that can be used in mobile AR services. Conventional methods could not have met all the requirement such as complexity, robustness and memory consumption for mobile AR services because of their trade-off relationship. In this paper, we propose a novel keypoint registration approach to solve the problem. Our method detects features from a small number of essential training images to improve robustness to changes in viewpoint. Experimental results demonstrate that our approach both improves the robustness of the conventional method by about 55% and speeds up runtime processing by about 7%.
