In this paper, we propose an end-to-end approach to performing multi-class, real-time detection, prior to the specific classes of motorcycles, helmets, and license plates in traffic monitoring. Using YOLO (You Only Look Once) and EasyOCR (Easy Optical Character Recognition), this system can be used to accurately detect and track motorcycles, helmet status and license plate in video streams. Extensive testing has shown the model to be capable of capturing a wide range of traffic movement and behavior characteristics (e.g., motorcycle counting, helmet compliance) Difficulty tracking little motorcycles (and consequently predicting their positions) is the main area of poor performance, especially when they are driving somewhat zigzag. More sophisticated tuning of the model algorithms and parameters are pending to adapt the model to mitigate these challenges and contribute to a better real-world performance. We will need to continue to collect and validate data to improve reliability and accuracy. In summary, our study helps to promote traffic monitoring technology, which provide useful tools for traffic management, safety enforcement and crash surveillance.
Real-Time Multi-class Helmet Violation Detection Using YOLOv8 with License Plate Recognition
Lect. Notes in Networks, Syst.
International Conference on Digital Technologies and Applications ; 2024 ; Benguerir, Morocco May 10, 2024 - May 11, 2024
2024-08-31
11 pages
Article/Chapter (Book)
Electronic Resource
English
Vehicle license plate recognition system for traffic violation detection
European Patent Office | 2015
|European Patent Office | 2022
|License plate recognition technology for toll violation enforcement
Tema Archive | 1990
|Violation enforcement system through rear license plate recognition
European Patent Office | 2024
|European Patent Office | 2015
|