Seite - 355 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

| | (3) 4.2. Position Estimation Range Narrowing Down For the narrowing down of the position estimation range, a value considering the estimation error is used as the estimated movement distance obtained by comparing the front and rear sensor images. In addition, self-evaluation equation (4) in UMap is used to select the reference position of position estimation to be performed again. P () is a self-evaluation value, A () is the number of nonzero pixels, is the i-th database image, SEN is the input line segment image, and LC is the logical product of and SEN. 5. Enhancement of Database using Wi-Fi Signals 5.1. Utilization of Wi-Fi In Wi-Fi-based fingerprinting, a client device obtains a set of Received Signal Strength Indication (RSSI) values for multiple Wi-Fi access points. The obtained set of RSSI values is used to look up the priorly created RSSI map. Therefore, a pre-learned map is also required for Wi-Fi fingerprinting. Although Wi-Fi fingerprinting is useful to some extent, the wireless environment fluctuates due to the movement of people and objets, which results in inaccurate localization. To cope with this problem, we use RSSI values only for deciding a rough location. The advantage of this approach is that we do not need a very accurate RSSI map beforehand. 5.2. Clustering As described above, we use RSSI values only for identifying a rough location of the device. We divide the localization target into areas based on RSSI values. Figure 4 illustrates how areas are defined. As some location, a client device obtains a set of RSSI values in addition to an image. The looked-upon image database is structured with area-based clusters. Therefore, the image is only searched in the appropriate area and thus lookup time can be reduced. One remaining issue is multiple similar images in the same area. This needs to be examined with regard to the size of an area. W.Takahashi etal. /UsingWi-FiSignals forCreatingaClustered ImageDatabase inUMap 355