Road traffic cccidents caused by speeding are one of the leading causes of disability and death around the world. The control of a large number of vehicles on the road needs to be supported by intelligent and automated traffic control systems to reduce human errors and increase accident reduction. Mostly, these systems monitor traffic using street cameras and identify illegal traffic behaviors, such as speeding violations. This paper proposes a smart camera system that utilizes Nano Pi M1 controller board which is embedded by deep learning YOLOv3 algorithm for object detection. The detection is applied for traffic information such as traffic jam including vehicle volume on road, vehicle type, and vehicle license plate; also applied for traffic violation detections including signal jump, vehicle speed. Furthermore, the proposed system is optimized for processing speed and has portability due to its small size and low energy consumption. The proposed smart camera system was tested in real time on highways and routes in Vietnam. The experimental results demonstrate the system’s superiority with an average frame processing time of 45 milliseconds and an accuracy of over 90% on 640×360 pixels resolution. It is superior in terms of computational resources, network bandwidth and investment costs for the system compared to previously studied systems.
Design Of Smart Camera System For Traffic Scene Understanding
2021-11-01
557410 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Real-Time Stereo Vision for Urban Traffic Scene Understanding
British Library Conference Proceedings | 2000
|