| Autor: |
Ronariv, Raien, Antonio, Renaldi, Jorgensen, Steven Farrelio, Achmad, Said, Sutoyo, Rhio |
| Předmět: |
|
| Zdroj: |
Procedia Computer Science; 2024, Vol. 245, p627-636, 10p |
| Abstrakt: |
Traffic congestion is an important problem that causes wasted gasoline, lost time, and stress. Traffic congestion is caused by the high number of motorized vehicles on the road and ineffective traffic light control. Dynamic traffic lights are a solution to reduce traffic congestion by changing the duration of the lights based on vehicle density. This research uses object detection algorithms for car tracking and finding the most effective algorithm. The result of this work can be utilized in real-time analysis of traffic conditions by detecting and tracking vehicles at crossroads. This work tests and compares Euclidean Distance Tracking and YOLO algorithms for tracking moving objects. The testing dataset is traffic videos from highway CCTV footage. The datasets were pre-processed and passed to the tracking algorithms. Lastly, the performance of both algorithms is compared. The experiment shows that the Euclidean algorithm performs worse than YOLO. Euclidean algorithm erroneously identifies certain parts of the vehicle, such as the windows, rear-view mirrors, and even the vehicle's shadows, as part of the vehicle itself. In contrast, YOLO can effectively track vehicles with minimal errors. This study also highlights limitations in accurately tracking and counting vehicles using the current system and video footage. For example, the tracking performance was affected in scenarios involving camera movement or multiple vehicle types. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
