Real-time sensor-fusion based indoor localization for mobile Augmented Reality

In this paper we propose a sensor fusion based indoor localization method for mobile Augmented Reality (MAR). The aim of this research is to provide fine-tuning of the line feature based localization to accuracy of centimeter-level by exploring multi-modality of a mobile device. In order to match with line features from the captured scene and the given floor map, a line-based indoor scene analysis is proposed with Manhattan world assumption. An efficient pairwise line matching method using corresponding compass sensor data is presented to yield accurate localization and registration for MAR. Experimental results demonstrated that the proposed method is able to provide real-time performance and robustness in indoor environment.