Very Fast Fingerprinting DB Construction for Precise Indoor Localization

Abstract

As the supply of smartphones has increased recently, demand for indoor localization has also increased. Fingerprinting technology is one of the representative indoor localization technologies. This localization technology compares the RF signal measurement value of user's location in indoor space with the fingerprinting DB. Therefore, the key to fingerprinting-based indoor localization is how accurately and quickly to construct the fingerprinting DB. Until now, it was time and cost-consuming to collect and calibrate RF measurements of the entire indoor spaces to construct the fingerprinting DB. To solve this problem, we propose indoor image-based generation algorithm of data collection path for the fast construction of fingerprinting DB. The proposed technology generates pixel-unit reference map data thorough image analysis of map features such as corridors, walls, and rooms displayed in RGB color on the map image. Based on the reference map data, the proposed technology corrects the path taken by the data collector using the map-matching algorithm, and the collected RF measurements are matched to each location. Finally, it constructs the fingerprinting DB through RSS propagation model so that the collected data is distributed not only on the path but also on the width like the corridor width. To verify the performance of the proposed technology, we have constructed the fingerprinting DB in a complex department store with 6 Bluetooth beacons. We measured the accuracy in the test bed using the previously proposed Surface Correlation-based fingerprinting technology. It took less than 30 minutes to construct the fingerprinting DB based on the collected data, and the localization accuracy was less than 3 meters. Through these results, we confirmed that the proposed technology can construct the fingerprinting DB accurately and quickly even in a real complex indoor environment.