Real-time traffic flow estimation based on Deep Learning using CCTV videos

Mrs Nilani Algiriyage1, Dr Raj  Prasanna1, Dr Kristin Stock2, Dr Emma Hudson-Doyle1, Prof David Johnston1

1Joint Centre for Disaster Research, Massey University, Wellington, New Zealand
2Institute of Natural and Mathematical Sciences, Massey University, Auckland, New Zealand

Identification of traffic flow is the first step that leads to the effective management of road traffic infrastructure and deployment of intelligent transportation systems. Nowadays, CCTV cameras are mounted in most of the places in cities, and they provide a massive amount of data in real-time. The data generated by CCTV cameras can be used as the foundation for accurate traffic flow estimation.

In this poster, we present our early work of a system estimating traffic flow in real-time using deep learning from CCTV video data. Therefore, a case is selected at one of the busiest roads in Christchurch Central Business District (CBD), New Zealand. CCTV video data is obtained from the New Zealand Transport Agency (NZTA). During the first stage, we have analysed around 150 frames, including more than 200 objects. You Only Look Once (Yolov3) algorithm is used to detect and classify vehicles. Also, we use Simple Online and Realtime Tracking (SORT) for vehicle tracking. Furthermore, a heuristic-based algorithm is introduced to count the vehicles by movement direction such as “left-lane” and “right-lane”. Our initial results show a mean absolute percentage error that is less than 12%.

Upon the completion of the system, city council authorities can use it to understand traffic flow patterns in real-time, make traffic predictions, understand anomalies, and make management decisions.


Biography:

Nilani is a PhD candidate at the Joint Centre for Disaster Research at Massey University, Wellington, New Zealand. She obtained her Masters Degree from Department of Computer Science and Engineering, University of Moratuwa in April 2015. Her Bachelor’s degree was in Management and Information Technology, graduating with a first class honors from University of Kelaniya in 2011.  Before she joined the Joint Centre for Disaster Research, she had been working as a lecturer in a government university, Sri Lanka.

About the conference

eResearch Australasia provides opportunities for delegates to engage, connect, and share their ideas and exemplars concerning new information centric research capabilities, and how information and communication technologies help researchers to collaborate, collect, manage, share, process, analyse, store, find, understand and re-use information.

Conference Managers

Please contact the team at Conference Design with any questions regarding the conference.

© 2018 - 2020 Conference Design Pty Ltd