Seite - 353 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

There are several works on clustering images. Pandey et al. [7] focus on the nature of images and make a hierarchy among images. In contrast, ours utilizes rough geolographical information for clustering. 3. Universal Map UMap is a position estimation infrastructure using a preliminary map with line segments as landmarks, and takes a map image database search method based on similarity calculation. Since maps are used, map construction like SLAM is not performed, so it can be said to be low computation. Also, this is a server client method, all processes are performed by the server, and the client performs camera shooting only, so it costs less. A schematic diagram of UMap is shown in Fig. 1. Umap consists of configuration server, clients, the three agents. Performing position estimation using the sensor image data received from a client on the server, and subsequently the client receives the estimated position. The agent detects a change in the environment, inform the server. The server performs a semi-permanent system operation by adding writes the changes to the database. Figure 1. Universal Map. 4. Design of Consistency Adjustment In UMap, it is not possible to estimate the position by taking time in consideration, and in a structural environment with high similarity, different position estimation may be done. Therefore, by comparing feature points of the input sensor image used in UMap, the existence position is stopped stochastically. After that, the position estimation range is restricted before and after the estimated position with the smallest error, and the position estimation is performed again to perform position estimation with temporal consistency. W.Takahashi etal. /UsingWi-FiSignals forCreatingaClustered ImageDatabase inUMap 353