Road traffic accident recognition is essential in providing timely information to healthcare authorities for reducing fatalities. This area of research is heavily dependent on traffic flow data captured through roadside sensing technologies that are installed on highways and intersections. To date, Deep Learning (DL) has achieved remarkable progress in solving time-series problems with increasing applications in road traffic accident recognition. This paper explores recent studies of DL techniques using roadside sensor-based traffic flow data. Limited literature has focused on road traffic accident recognition in mixed traffic. Various issues in current DL recognition solutions that affect accuracy, including consideration of user varieties, dynamic traffic flow conditions, and external environmental factors are discussed. In this research, a fusion feature-based deep learning model for traffic accident recognition has been proposed, consisting of three major streams of models to cater for prominent features in traffic accident recognition in a mixed traffic flow environment.
Exploring Deep Learning in Road Traffic Accident Recognition for Roadside Sensing Technologies
Lect. Notes in Networks, Syst.
Proceedings of the Future Technologies Conference ; 2022 ; Vancouver, BC, Canada October 20, 2022 - October 21, 2022
Proceedings of the Future Technologies Conference (FTC) 2022, Volume 1 ; Kapitel : 3 ; 32-52
2022-10-13
21 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Road traffic accident risk prediction deep learning algorithm
Europäisches Patentamt | 2022
|Roadside sensing equipment and intelligent traffic system
Europäisches Patentamt | 2021
|