© 2021 IIETA. This article is published by IIETA and is licensed under the CC BY 4.0 license (http://creativecommons.org/licenses/by/4.0/).
OPEN ACCESS
The utilization and the quality of public transport are important for the customers, maintainers and service providers. Passive measurement techniques, when humans are not involved are the cheapest way for collecting large amounts of long-term data from multiple public transport lines. Useful data can be collected from various sources, such as from cameras, infrared sensors and Wi-Fi routers. We addressed the problems of estimating passenger counts in two different ways, and also to get travel statistics like the number of passengers getting on or off a vehicle at a bus stop; and even to compute an origin–destination matrix from Wi-Fi monitoring data. In this study, we focus on Wi-Fi data, which can be still useful for extracting relevant data after many years. here we describe Wi-fi data collection methods, and then prove the usefulness of applying simple artificial intelligence-based methods to extract information from the huge amount of Wi-Fi data. We will also show that ‘lower-level re-estimation’ can be useful for further optimization, which means that globally modelled data may have to be re-modelled on partially selected groups to get better results. Namely, after building linear models and estimating absolute and relative errors, we found that the relative error of the Wi-Fi-based estimation can be markedly reduced if data are processed and analysed in more detail. When a daily Wi-Fi analysis is split into between-stops parts, an additive linear correction can be computed and applied to these parts, and as a result, the relative error of estimates can be reduced.
axle load-based estimation, public transport, Wi-Fi frame monitoring
[1] N ielsen, B. F., Frølich, L., Nielsen, O. A., & Filges, D. (2014). Estimating passengernumbers in trains using existing weighing capabilities. Transportmetrica A: TransportScience, 10(6), 502-517.
[2] K otz, A. J., Kittelson, D. B., & Northrop, W. F. (2015). Novel vehicle mass-basedautomated passenger counter for transit applications. Transportation Research Record,2563(1), 37-43.
[3] H su, Y. W., Wang, T. Y., & Perng, J. W. (2020). Passenger flow counting in buses basedon deep learning using surveillance video. Optik, 202, 163675.
[4] Erlik Nowruzi, F., El Ahmar, W. A., Laganiere, R., & Ghods, A. H. (2019). In-vehicleoccupancy detection with convolutional networks on thermal images. Proceedings ofthe IEEE Conference on Computer Vision and Pattern Recognition Workshops (pp. 0-0).
[5] B alasubramanian, Y., Chandrasekaran, N., Asokan, S., & Subramanian, S. S. (2019).Deep-facial feature-based person reidentification for authentication in surveillanceapplications. Visual Object Tracking in the Deep Neural Networks Era. IntechOpen.
[6] L i, T., Sun, D., Jing, P., & Yang, K. (2018). Smart card data mining of public transportdestination: a literature review. Information, 9(1), 18.
[7] K urkcu, A., & Ozbay, K. (2017). Estimating pedestrian densities, wait times, and flowswith Wi-Fi and Bluetooth sensors. Transportation Research Record, 2644(1), 72-82.
[8] C hilipirea, C., Baratchi, M., Dobre, C., & Steen, M. V. (2018). Identifying stops andmoves in WiFi tracking data. Sensors, 18(11), 4039.
[9] M yrvoll, T. A., Håkegård, J. E., Matsui, T., & Septier, F. (2017, October). Countingpublic transport passenger using WiFi signatures of mobile devices. 2017 IEEE 20thInternational Conference on Intelligent Transportation Systems (ITSC) (pp. 623-628).IEEE.
[10] O ransirikul, T., Nishide, R., Piumarta, I., & Takada, H. (2014). Measuring bus passengerload by monitoring Wi-Fi transmissions from mobile devices. Procedia Technology, 18,120-125.
[11] K ang, L., Qi, B., & Banerjee, S. (2016, February). A wireless-based approach fortransit analytics. Proceedings of the 17th International Workshop on Mobile ComputingSystems and Applications (pp. 75-80).Table 3: C orrelations between passenger count estimates.Correlation, mass-based vs. human counted 0.97 [1] 0.99 [2] 0.98 (ours)Correlation, Wi-Fi based vs. human counted orload based (*)0.62 [10] 0.76 [11] 0.8 (ours*)András Bánhalmi et al., Int. J. Transp. Dev. Integr., Vol. 5, No. 1 (2021) 27
[12] R eview of the TfL WiFi pilot, available online: http://content.tfl.gov.uk/review-tfl-wifipilot.pdf
[13] N guyen, K. A., Wang, Y., Li, G., Luo, Z., & Watkins, C. (2019). Realtime tracking ofpassengers on the London underground transport by matching smartphone accelerometerfootprints. Sensors, 19(19), 4184.
[14] F reudiger, J. (2015, June). How talkative is your mobile device? An experimental studyof Wi-Fi probe requests. Proceedings of the 8th ACM Conference on Security & Privacyin Wireless and Mobile Networks (pp. 1-6).
[15] M artin, J., Mayberry, T., Donahue, C., Foppe, L., Brown, L., Riggins, C. & Brown, D.(2017). A study of MAC address randomization in mobile devices and when it fails.Proceedings on Privacy Enhancing Technologies, 2017(4), 365-383.
[16] C astignani, G., Montavont, N., & Arcia-Moret, A. (2010). Analysis and evaluation ofWiFi scanning strategies. Proceeding of IV Cibelec, 3-7.