Research On Fingerprint Crowdsourcing-based Indoor Localization

Motivated by the Internet and artificial intelligence, there is a great increase of the requirements of location-based service in indoor environment. For emergency rescue or applications in everyday life such as navigation, personalized recommendation, people need to obtain their locations in indoor environment. Indoor localization technique has been studied for more than two decades. Although a variety of indoor localization techniques have been proposed, including triangulation, fingerprinting, distance-based positioning, dead-reckoning and etc., none of them can fulfill the requirement of both deployment costs and localization accuracy in commerce. Indoor localization system is still a research focus.This thesis adopts WiFi fingerprinting technique with relative lower deployment costs and higher localization accuracy. Traditional WiFi fingerprinting requires labor-intensive site survey to construct fingerprint database, which is one of the bottlenecks of the fingerprinting technique. This thesis focuses on fingerprint crowdsourcing-based localization to avoid the labor-intensive site survey. First, we analyze crowdsourcing-based indoor localization systems and identify two crucial problems: crowdsourcing fingerprint annotation and diversity device calibration. The classification is based on the pattern of crowdsourcing, diversity device calibration and localization accuracy. We then propose a crowdsourcing fingerprint-based indoor localization system aiming to solve the two crucial problems. The innovation of the proposed system is to utilize the structure of fingerprint to extract subarea fingerprints from crowdsourcing and to realize the subarea localization of mobile devices.This thesis designs and realizes the proposed system, and has conducted field experiments in three different real indoor environments. Our experiment results show that in typical indoor environments, the proposed scheme can achieve over 95% hitting rate to correctly locate a mobile device to its subarea. And the proposed online positioning algorithm can help to relieve the device diversity problem in terms of improved localization accuracy, compared with the classical nearest neighbor algorithm.