Parked Car Detection Method Based on Home Wireless LAN -Using Household Equipment to Detect a Parked Car-

Abstract

Consumer-to-consumer (C2C) car park sharing services are now widespread. In these services, space owners rent their unused land, e.g., an unused parking space next to the owner’s home, to a driver. Different from coin parking services, drivers using the park sharing services can make reservations through web sites or mobile-phone applications. There are drivers who abuse this system by, for example, overstaying their reservation. The high cost of parking space sensors that monitor the existence of a parked car makes it cost prohibitive for individuals to install these types of equipment to deter the abuse of parking spaces. To address this problem, we propose a parked car detection technique employing wireless LAN. The proposed method utilizes a household wireless LAN access point and a household laptop computer, and we monitor the received signal strength indicator (RSSI) of beacon frames. This unique method reduces the detection cost and will aid in proliferating car park sharing services. In this paper, we detect the existence of a car in a parking space using the mean and variance of the RSSI. The generated datasets are used for Support Vector Machine (SVM), Decision Tree, and the K-nearest neighbor algorithm (KNN) binary classification. We previously showed that a parked car can be detected when the locations of the equipment are the same. In this study, we show the effect of changing the equipment location in the house on the detection score. These results show that the best recall score is 0.88 and that changing the equipment location does not affect the recall score. The results of this study are expected to be used in a multimodal system.