A Deep Learning Based System for Intelligent Traffic Monitoring

Mostafa  Mah; Zohre Arabi; Alireza  Sehat

doi:10.48314/des.vi.82

نویسندگان

Mostafa Mah * Department of Computer Engineering and Information TechnologyPayame Noor UniversityTehran, Iran
Zohre Arabi Department of Computer Engineering, Technical and Vocational University (TVU), Tehran, Iran
Alireza Sehat Department of Computer Engineering and Information TechnologyPayame Noor UniversityTehran, Iran

https://doi.org/10.48314/des.vi.82

چکیده

Traffic monitoring and analysis are critical for efficient urban transportation management and safety. In this paper, we present a deep learning–based system capable of real-time detection, tracking, counting, and speed estimation of vehicles from video streams. The system integrates the YOLOv9 object detection model with the Deep Sort multi-object tracker to maintain consistent identities of vehicles across frames. To estimate speed, we employ a perspective transformation approach that maps pixel coordinates to real-world distances, enabling approximate calculation of vehicle velocities in km/h. The implementation is further enhanced by a user-friendly web interface built with Streamlit, providing live video visualization, frame-per-second (FPS) metrics, vehicle count statistics, average speed, and class-wise distributions via interactive charts. Experimental results demonstrate that the system can efficiently monitor traffic flow while offering actionable insights for urban traffic analysis, transportation planning, and real-time surveillance applications. This work highlights a practical approach to combining state-of-the-art object detection and tracking techniques with accessible visualization tools for intelligent traffic monitoring.