An AI-Powered Traffic Surveillance Framework for Vehicle Tracking and Automated Violation Detection using YOLOv10, Computer Vision, and Deep Neural Networks

Abstract

The rapid growth of urbanization and the increasing number of vehicles have made traffic monitoring and law enforcement more challenging. Conventional traffic surveillance systems rely heavily on manual observation or passive video recording, resulting in delayed violation detection, increased human effort, and reduced monitoring efficiency. To address these limitations, this research proposes an AI-powered traffic surveillance framework that integrates YOLOv10, computer vision, and deep neural networks for real-time vehicle detection, tracking, and automated traffic violation detection. The proposed system is designed to improve road safety by accurately identifying traffic violations such as signal jumping, lane violations, overspeeding, and illegal parking while supporting continuous surveillance in complex traffic environments. The proposed framework follows a modular architecture consisting of video acquisition, image preprocessing, vehicle detection using YOLOv10, multi-object tracking with DeepSORT/Centroid Tracking, violation analysis, and alert generation. OpenCV is employed for real-time video processing, while deep learning techniques enable accurate vehicle classification and localization. The tracking module maintains unique identities for vehicles across consecutive frames, allowing reliable analysis of vehicle trajectories and behavior. A Flask-based web interface provides live visualization of detected vehicles, tracking IDs, violation alerts, and automated storage of violation records for future analysis and reporting. Experimental implementation demonstrates that the proposed framework achieves efficient real-time vehicle detection and tracking with high accuracy under normal traffic conditions while significantly reducing manual monitoring efforts. The integration of deep learning-based object detection and intelligent tracking improves the reliability of automated traffic surveillance and supports scalable deployment in smart transportation systems. The proposed solution offers an effective platform for intelligent traffic management and can be further enhanced by incorporating Automatic Number Plate Recognition (ANPR), advanced transformer-based detection models, edge computing, cloud integration, and AI-driven traffic analytics for next-generation smart city applications.