Points of interest detection for map-aided PDR in combined outdoor-indoor spaces

摘要: Complementary data are necessary to bind the positioning error growth of Pedestrian Dead Reckoning (PDR). In this paper, absolute position updates made possible with online detection different types points interest (POIs) located on map. The POIs selected depending specific motion patterns which associated locations To create database, correlation between pedestrian and map is first studied outcome a typology POIs. A K-NN (K nearest neighbors) algorithm used train modes, further exploited for in order update PDR data. Experimental assessment POI-based calibration conducted both outdoor indoor spaces focus transition environments. 90% time, correctly classified corrected an accuracy that depends features (width corridor/door, staircase size…). Therefore, approach found be promising enhancing using only