Seite - 352 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

Using Wi-Fi Signals for Creating a Clustered Image Database in UMap Wataro TAKAHASHI a,1 , Junji TAKAHASHI b and Yoshito Tobe a a Department of Integrated Information, Aoyama Gakuin University, Kanagawa, Japan b Department of Mechanical Engineering, Kagoshima University, Kagoshima, Japan Abstract. UMap is an indoor localization system based on the image-based similarity between a photograph taken at a client smartphone and the database at the server. A problem in the current UMAP is the inefficiency in searching the appropriate image in the database. In this paper, we propose reorganization of the database by clustering the registered images based on the rough location estimated by Wi-Fi signals. This paper describes the design of the enhanced UMap database system. Keywords. localization, line-segment-feature, 2D-3D matching, Wi-Fi, database 1. Introduction Towards realizing intelligent environments, providing a high precision indoor localization system is necessary. Since the signals from Global Navigation Satellite System (GNSS) do not always penetrate the structure of building, we cannot rely on GNSS inside buildings. There are sevel alternatives to GNSS in indoor position estimation: Wi-Fi Fingerprint [1] and SLAM (Simultaneous Localization and Mapping) [2-3]. However, their installation and creation costs are high because they use infrastructure and expensive sensors. Based on the background, we have developed a low-computation, low-cost location estimation infrastructure that uses an image called UMap (Universal Map) [4]. UMap relies on the central database containing images taken by clients. A problem in the database is its scalability; the average search time of the target image increases lineary as the number of registered images. In this work, we propose reorganization of the database by clustering the registered images based on the rough location estimated by Wi-Fi signals. This paper describes the design of the enhanced UMap database system. 2. Related Work There is a problem that the estimation result differs in position estimation, and research to solve it has been done. Grisetti and colleagues proposed a method called Graph-Based SLAM that uses sensor information such as robot odometry to narrow down the position of robot stochastically and estimate the position [5]. In this research, we refer to the theory of Graph-Based SLAM and make it possible to estimate position with time consistency. 1 Wataro TAKAHASHI, Department of Integrated Information Technology, Aoyama Gakuin University, 5-10-1 Fuchinobe, Chuo-ku, Sagamihara 252-5258, Japan E-mail: wataro@rcl-aoyama.jp Intelligent Environments 2019 A. Muñoz et al. (Eds.) © 2019 The authors and IOS Press. This article is published online with Open Access by IOS Press and distributed under the terms of the Creative Commons Attribution Non-Commercial License 4.0 (CC BY-NC 4.0). doi:10.3233/AISE190065 352