Edge Computing for Traffic Characterization: History
View Latest Version
Please note this is an old version of this entry, which may differ significantly from the current revision.
Subjects: Engineering, Civil
Contributor:
Asif Khan
,
Khurram S. Khattak
,
Zawar H. Khan
,
Thomas Aaron Gulliver
,
Zakhil Abdullah

Traffic flow analysis is essential to develop smart urban mobility solutions. Although numerous tools have been proposed, they employ only a small number of parameters. To overcome this limitation, an edge computing solution is proposed based on nine traffic parameters, namely, vehicle count, direction, speed, and type, flow, peak hour factor, density, time headway, and distance headway.

  • edge computing
  • traffic monitoring
  • urban mobility
  • Internet of Things

1. Introduction

Urbanization is expected to drive global economic growth in the coming decades by increasing productivity and reducing poverty. However, this growth is threatened by challenges to urban mobility including greenhouse gas (GHG) emissions and lost productivity due to road accidents and congestion. The road transportation sector accounts for 25% of worldwide fuel consumption and 29% of GHG emissions [1]. Traffic congestion also results in a significant reduction in productivity, with an average driver in the USA losing 36 h and USD 564 in 2021 [2]. Moreover, according to the World Health Organization (WHO), road accidents cause 1.3 million deaths and 50 million non-fatal injuries every year [3]. Therefore, it is imperative to develop innovative and effective solutions such as intelligent transportation systems (ITS) to mitigate these challenges and ensure efficient urban mobility.
ITS-based solutions are a promising means of improving road network efficiency. Detailed traffic data including vehicle count, speed, and classification, flow, spatial/temporal densities, vertical/horizontal headways, road capacity, heatmaps, and trajectories are essential to provide insights for traffic engineers to improve transport network management. Furthermore, these parameters can be employed in traffic simulation software [4][5] to aid urban planners in designing effective road networks.
Both intrusive and non-intrusive traffic monitoring systems have been developed. However, these solutions have limitations, including only measuring traffic count and speed, installation and maintenance difficulties, and high costs [6]. With advancements in image processing techniques, roadside video can now be employed. While Internet-of-Video-Things (IoVT) solutions are effective, the high bandwidth requirements for roadside video transmission to servers are a major limitation [6][7]. Image processing edge computing solutions have been proposed to overcome this problem. However, the computational power of devices such as Raspberry Pi (RPi) limits the ability to provide detailed traffic information. Existing edge computing solutions provide either count [1][8][9][10], count and speed [11][12][13][14], or count and classification [11][15][16][17][18].
An edge computing solution is proposed to overcome the limitations of existing traffic monitoring systems. The objective is to accurately obtain vehicle count, speed, type, and direction, flow, peak hour factor, density, time headway, and distance headway. This is achieved using a sensor node composed of an RPi 4, Pi camera, Intel Movidius Neural Compute Stick 2, Xiaomi MI 10,000 mAh power bank, and Zong 4G Bolt+. The pre-trained Mobilenet-SSD model from the Intel OpenVINO Toolkit is employed for vehicle count and classification [19]. Vehicle speed is estimated using a centroid tracking algorithm running on the RPi 4. The measured traffic parameters are transmitted to the ThingSpeak cloud platform using 4G. This data can be used for traffic flow analysis to aid in transportation network planning and management.

2. Count Solutions

An edge computing solution based on an RPi and web camera was presented in [9] which achieves a vehicle count accuracy of 83%. Vehicle count, road density, time headway, and vehicle emissions were obtained with the system in [1]. This solution used an RPi 4 and Pi camera with four sensors to measure carbon monoxide, carbon dioxide, and particulate matter. A vehicle count accuracy of 86% was reported and the measured parameters were transmitted to the ThingSpeak cloud platform using the RPi Wi-Fi module.

3. Count and Classification Solutions

In [15], a system to count and classify vehicles at a highway toll booth was developed using an RPi B and Pi camera [8]. In [10], an edge computing solution to count and classify vehicles was presented which employs an RPi 2, Pi camera, and MySQL web server database [17]. In [16], an RPi B and Samsung smart security camera were used to transmit parameters to a remote web server for display. A vehicle count accuracy of 83% was obtained. In [18], an edge computing solution was developed using an RPi 2 and Pi camera to count vehicles and classify them as small or large. The data were stored locally on the RPi 2 for archiving purposes. In [20], a real-time stereo vision system was presented to count vehicles and classify them as cars or small or big trucks. It employs an RPi 3B and USB webcam and transmits the parameters to a local web server for display.

4. Count and Speed Solutions

A solution using an RPi 2B+ and Pi camera to count vehicles and estimate their speed was given in [12]. The Flask web framework was used to archive the parameters on an edge cloud server. In [21], an edge computing solution to count vehicles and estimate their speed was presented which uses an RPi 2 and Pi camera. The effect of different frame sizes on the CPU and memory was examined. It was found that the CPU performance was not significantly affected by the frame size, but higher-resolution frames required more memory. In [22], a system to count vehicles and estimate their speed was developed which uses an RPi 3B and Pi camera. The parameters were stored locally on the RPi, and a count accuracy of 100% and speed accuracy of 90% were reported. The solution proposed in [13] uses an RPi 3B and Pi camera. All of these edge computing solutions are limited by the RPi computing resources. Thus, an edge computing solution is proposed here to overcome this constraint. The advantages of this solution are as follows.
  • Existing edge computing solutions only measure two traffic parameters, either vehicle count and classification [15][16][17][18][20][23] or vehicle count and speed [5][12][13][21][22]. Conversely, the proposed solution can measure nine traffic parameters, namely, vehicle count, speed, direction, and type, flow, peak hour factor, density, time headway, and distance headway.
  • The proposed solution can classify five different types of vehicles, namely, cars, buses, motorbikes, bicycles, and animal-drawn carts (horse and cow). This is greater than the number of vehicle classes provided by existing solutions [15][16][17][18][20].
  • The proposed solution can count and classify vehicles with an accuracy of 93%, which is better than the accuracy reported in previous studies [1][4][9][14][16][22].
  • The proposed solution can count and estimate the speed of a wide variety of vehicles as well as pedestrians. This includes trams (trains), airplanes, and boats. This is because the detection model was trained on over 70 different objects, including these vehicles. Vehicle direction is also obtained. This makes the proposed system ideal for characterizing heterogeneous traffic behavior. Note that no other system provides the direction of vehicles.
  • The proposed solution was designed considering cost, reliability, and scalability. The sensor node costs less than USD 300 and has a low power consumption of 1.2 A per hour. Unlike previous systems, the proposed solution transmits the measured parameters to a cloud platform using 4G with a data bandwidth requirement of approximately 1.5 MB per hour.

This entry is adapted from the peer-reviewed paper 10.3390/s23239385

References

  1. Khan, A.; Khattak, K.S.; Khan, Z.H.; Khan, M.A.; Minallah, N. Cyber physical system for vehicle counting and emission monitoring. Int. J. Adv. Comput. Res. 2020, 10, 181–193.
  2. Texas A&M Transportation Institute. 2021 Urban Mobility Report. Available online: https://mobility.tamu.edu/umr/ (accessed on 21 June 2023).
  3. World Health Organization. Occupant Restraints: A Road Safety Manual for Decision-Makers and Practitioners. Available online: https://www.who.int/publications/m/item/occupant-restraints--a-road-safety-manual-for-decision-makers-and-practitioners (accessed on 21 June 2023).
  4. Bowman, C.N.; Miller, J.A. Modeling traffic flow using simulation and big data analytics. In Proceedings of the Winter Simulation Conference, Washington, DC, USA, 11–14 December 2016; pp. 1206–1217.
  5. Zeb, A.; Khattak, K.S.; Rehmat Ullah, M.; Khan, Z.H.; Gulliver, T.A. HetroTraffSim: A macroscopic heterogeneous traffic flow simulator for road bottlenecks. Future Transp. 2023, 3, 368–383.
  6. Ayaz, S.; Khattak, K.; Khan, Z.; Minallah, N.; Khan, M.; Khan, A. Sensing technologies for traffic flow characterization: From heterogeneous traffic perspective. J. Appl. Eng. Sci. 2022, 20, 29–40.
  7. PROTOTXT File Extension (2021) PROTOTXT File—What is a .Prototxt File and How Do I Open It? Available online: https://fileinfo.com/extension/prototxt (accessed on 16 August 2023).
  8. Gupta, A.; Gandhi, C.; Katara, V.; Brar, S. Real-time video monitoring of vehicular traffic and adaptive signal change using Raspberry Pi. In Proceedings of the IEEE Students’ Conference on Engineering & Systems, Prayagraj, India, 10–12 July 2020.
  9. Prabhu, S.; Kawale, S.; Mallampati, M.; Dillip, M.; Sahu, N.; Huluka, M. Automated vehicle detection and tracking using Raspberry-Pi. Int. J. Mech. Eng. Mechatron. 2022, 7, 66–72.
  10. Ramesh, R.; Divya, G.; Dorairangaswamy, M.A.; Unnikrishnan, K.N.; Joseph, A.; Vijayakumar, A.; Mani, A. Real-time vehicular traffic analysis using big data processing and I0T based devices for future policy predictions in smart transportation. In Proceedings of the International Conference on Communication and Electronics Systems, Coimbatore, India, 17–19 July 2019; pp. 1482–1488.
  11. Zakhil, A.; Khan, A.; Khattak, K.S.; Khan, Z.H.; Qazi, A. Vehicular flow characterization: An Internet of video things-based solution. Pak. J. Eng. Technol. 2023, 6, 13–22.
  12. Jiménez, A.; García-Díaz, V.; Anzola, J. Design of a system for vehicle traffic estimation for applications on IoT. In Proceedings of the Multidisciplinary International Social Networks Conference, Bangkok, Thailand, 21–23 December 2017.
  13. Paglinawan, C.C.; Yumang, A.N.; Andrada, L.C.M.; Garcia, E.C.; Hernandez, J.M.F. Optimization of vehicle speed calculation on Raspberry Pi using sparse random projection. In Proceedings of the IEEE International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment and Management, Baguio City, Philippines, 29 November–2 December 2018.
  14. Kulkarni, A.P.; Baligar, V.P. Real time vehicle detection, tracking and counting using Raspberry-Pi. In Proceedings of the International Conference on Innovative Mechanisms for Industry Applications, Bangalore, India, 5–7 March 2020.
  15. Suryatali, A.; Dharmadhikari, V.B. Computer vision based vehicle detection for toll collection system using embedded Linux. In Proceedings of the International Conference on Circuits, Power and Computing Technologies, Nagercoil, India, 19–20 March 2015.
  16. Espinoza, F.T.; Gabriel, B.G.; Barros, M.J. Computer vision classifier and platform for automatic counting: More than cars. In Proceedings of the IEEE Ecuador Technical Chapters Meeting, Salinas, Ecuador, 16–20 October 2017.
  17. Gregor, D.; Ciekel, K.; Arzamendia, M.; Gregor, R. Design and implementation of a counting and differentiation system for vehicles through video processing. World Acad. Sci. Eng. Technol. Int. J. Comput. Inf. Eng. 2016, 10, 1771–1778.
  18. Sorwar, T.; Azad, S.B.; Hussain, S.R.; Mahmood, A.I. Real-time vehicle monitoring for traffic surveillance and adaptive change detection using Raspberry Pi camera module. In Proceedings of the IEEE Region 10 Humanitarian Technology Conference, Dhaka, Bangladesh, 21–23 December 2017; pp. 481–484.
  19. OpenVINO. mobilenet-ssd. Available online: https://docs.openvino.ai/2021.2/omz_models_public_mobilenet_ssd_mobilenet_ssd.html (accessed on 21 June 2023).
  20. Iszaidy, I.; Alias, A.; Ngadiran, R.; Ahmad, R.B.; Jais, M.I.; Shuhaizar, D. Video size comparison for embedded vehicle speed detection; travel time estimation system by using Raspberry Pi. In Proceedings of the International Conference on Robotics, Automation and Sciences, Melaka, Malaysia, 5–6 November 2016.
  21. McQueen, R. Detection and Speed Estimation of Vehicles Using Resource Constrained Embedded Devices; University of British Columbia: Kelownam, BC, Canada, 2018.
  22. Ullah, R.; Khattak, K.; Khan, Z.H.; Ahmad Khan, M.; Minallah, N.; Khan, A. Vehicular traffic simulation software: A systematic comparative analysis. Pak. J. Eng. Technol. 2021, 4, 66–78.
  23. Tsagkatakis, G.; Savakis, A. Random projections for face detection under resource constraints. In Proceedings of the IEEE International Conference on Image Processing, Cairo, Egypt, 7–10 November 2009; pp. 1233–1236.
More
© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.
This entry is offline, you can click here to edit this entry!
ScholarVision Creations
Feedback