Enhancing CCTV-Based Traffic Violation Detection Using AI for Automated Enforcement

The rapid increase in urban traffic violations has exposed the limitations of traditional CCTV-based surveillance systems, which are typically confined to detecting overspeeding or red-light breaches and rely heavily on manual monitoring. These systems lack the intelligence to automatically identify more complex violations such as helmet non-compliance or wrong-way driving. To overcome these challenges, we present an AI-enabled traffic violation detection framework that integrates YOLO, Faster R-CNN, and Deep SORT for real-time analysis. Faster R-CNN is used for detecting helmet usage on two-wheelers, outperforming YOLO with 94% accuracy and 91% mean Average Precision (mAP). YOLO is applied to detect other violations such as red light and stop-line breaches, wrong-lane driving, and anticipates vehicle trajectories, achieving 92% accuracy and 89% mAP. Additionally, the Deep SORT algorithm tracks vehicle directions and movements, ensuring precise detection of wrong-way driving. A Streamlit-based web interface enables automated video analysis and generates structured violation logs with annotated snapshots. Evaluations on real-world CCTV footage demonstrate the robustness and scalability of the solution for deployment in smart traffic environments.