| Autor: |
Rasheed, Raffik, Puma, Julio Cesar Tisnado, Kulothungan, Kishore, Kathirvelu, Bhaskar, Subramanian, Sendilvelan, Tripathi, Vikas |
| Zdroj: |
AIP Conference Proceedings; 2024, Vol. 2937 Issue 1, p1-7, 7p |
| Abstrakt: |
In modern era, road traffic density has been increasing requiring a smart traffic system in managing traffic. Traffic congestion management has been a challenge nowadays as there is sharp increase in the number of vehicles. Real-time vehicle monitoring techniques play an important part in prevention of recurring traffic jams, managing traffic on roads, and deadly accidents. The most notable challenges are operating the system in real time to accurately identify and classify vehicles in active traffic flows without obstructing the vehicle. Factors such as weather environment and video quality can become objectives in vehicle detection using traditional machine learning algorithms leading to poor detection results. Recently, for better application in traffic control systems for vehicles, there has been a move towards deep learning architectures. In this paper, YOLOv2 (You Only Look Once) principle is utilized for real-time detection and counting of vehicles. The working of proposed strategy is carried out in two phases where vehicle detection is done at first phase and then counting of moving vehicles during second phase. This method provides strong object positioning functionality and high frames per second (FPS). The test results confirm that higher detection accuracy can be achieved with the proposed method, even when detecting small vehicles. This mechanism of YOLO based system provided an average detection of 95.75% and counting accuracy of 96.85%. The implications of these results help in intelligent transport planning. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
