Automated traffic congestion estimation via public video feeds
Ricardo Sciberras, Frankie Inguanez · 2019
Route planning is essential for commuters and whilst many developed companies and urban cities have reliable systems certain areas in the world have to cope with inaccurate systems or the lack of. There is a current surge in public video feeds that are directed on key roads with some being subsidies by governments. Video footage for two arterial and two rural roads have been gathered for 6 days with a total of 154 hours of footage together with video recordings of the traffic congestion level for the same roads as estimated by an industry leading traffic application. In this research, we are presenting a solution that given such a public video feed will automatically provide traffic congestion levels after a 24 hour self-calibration period. We have benchmarked our solution with an industry leading traffic application and have obtained a 16% higher precision and an overall accuracy of 86%. This research is based on image processing techniques and has addressed clear weather conditions during the day and recommend further research into the estimation traffic during challenging scenarios.