Design of New Traffic System YOLO-LIO: Light-Traffic Intercept and Observation

Vehicle detection is an area of active development aimed at enhancing driving safety and ensuring compliance with traffic regulations. Despite ongoing efforts, accidents and traffic violations continue to pose significant challenges, leading to disruptions in driving. In response to these issues, the author aims to develop a more efficient traffic management system to improve driver organization and driving behavior. To achieve this, the author proposes using YOLO-LIO as the neural network of choice for the Traffic System. The effectiveness of YOLO-LIO was evaluated using three datasets: the Montevideo Audio and Video Dataset (MAVD), the GARM Road-Traffic Monitoring dataset (GRAM-RTM), and a custom dataset created by the author. The results highlight the superior performance of YOLO-LIO in vehicle detection tasks, achieving accuracy rates of $\mathbf{9 9. 0 2 \%}$ on the MAVD dataset, $\mathbf{9 9. 5 5 \%}$ on the GRAM-RTM dataset, and $\mathbf{9 9. 3 2 \%}$ on the custom dataset. This demonstrates the model’s high effectiveness across various datasets. Additionally, the author conducted experiments incorporating OCR technology with the YOLO-LIO algorithm in the Traffic System. The system achieved an accuracy of $\mathbf{8 0. 2 1 \%}$ in vehicle number detection, demonstrating its effectiveness. This result reflects the overall performance of the entire system process, from data input to the final detection output, ensuring a comprehensive and accurate detection mechanism. Compared to other algorithms such as YOLOv3 + OCR, YOLOv4, and Faster R-CNN, YOLO-LIO + OCR, they have exhibited significantly better performance. These promising results highlight the potential of YOLO-LIO in creating a robust Traffic System that can significantly enhance road safety and traffic regulation compliance.