Recently, the map matching-assisted positioning method based on micro-electromechanical systems (MEMS) inertial devices has become a research hotspot for indoor pedestrian positioning; however, these are based on existing indoor electronic maps. In this paper, without prior knowledge of the map and through building an indoor main path feature point map combined with the simultaneous localization and map building (SLAM) particle filter (PF-SLAM) algorithm idea, a PF-SLAM indoor pedestrian location algorithm based on a feature point map was proposed through the inertial measurement unit to improve indoor pedestrian positioning accuracy. Aiming at the problem of inaccurate heading angle estimation in the pedestrian dead reckoning (PDR) algorithm, a turn-straight-state threshold detection method was proposed that corrected the difference of the heading angles during the straight-line walking of pedestrians to suppress the error accumulation of the heading angle. Aiming at the particles that are severely divergent at the corners, a feature point matching algorithm was proposed to correct the pedestrian position error. Furthermore, the turning point extracted the main path that failed to match the current feature point map as a new feature point was added to update the map. Through the mutual modification of SLAM and an inertial navigation system (INS) the long-time, high-precision, and low-cost positioning functions of indoor pedestrians were realized.
Keywords: feature point matching; indoor localization; inertial navigation system (INS); particle filtering; simultaneous localization and map building (SLAM) algorithm.