Fusion of inertial and visual information for indoor localisation

Abstract

Indoor localisation has attracted a lot of attention because of its importance for location-based services. A fusion algorithm (named as YELM-DS) based on extreme learning machine (ELM) and Dempster–Shafer (D–S) evidence theory is proposed. ELM learns the input data model composed of inertial and visual information and target output positions with high speed. During online phase, the final localisation result of a frame is decided by the trust degree obtained from D–S. Angle judgments are also introduced to decrease the big localisation errors of turning. Compared with the existing vision-only methods, the proposed method can both run in real time and achieve good localisation accuracy even in challenging scenarios.