With the development of the intelligent transportation system, we propose a traffic accident detection system based on roadside traffic monitoring cameras. By combing deep learning-based object detection and tracking technology and expert system-based traffic accident identification technology, a stable traffic accident detection system is implemented in conventional scenarios. In detail, we create a dataset containing 30,000 images, the objects in which are labeled as accident participants, such as pedestrian, bicyclist, tricyclist, car, SUV, bus, and truck. Then the dataset is used to train an object detection and tracking network. Thirdly, we design the rules of accident identification based on spatial temporal constraints. Finally, the system was verified using 40 real road traffic accidents. The experimental results show that the system can achieve 90.91% precision and 81.08% recall.
Intelligent Traffic Accident Detection System Using Surveillance Video
Lect. Notes Electrical Eng.
2022-01-13
11 pages
Article/Chapter (Book)
Electronic Resource
English
Intelligent Traffic Accident Detection System Using Surveillance Video
British Library Conference Proceedings | 2022
|European Patent Office | 2022
|