Dehradun Vision 2025: Smart City Traffic Optimization Using YOLOv11, U-Net, and Deep SORT

This research presents an AI-powered traffic management system developed to solve Dehradun's urban mobility issues in alignment with Dehradun Smart City Vision 2025. It implements YOLOv11 for vehicle detection, U-Net for lane segmentation, Deep SORT for multi object tracking, and IoT-based adaptive signal control. Unlike fixed-timer systems, it adjusts signals in real-time considering traffic density, pedestrian movement, and other environmental factors. Using 3000 images from eight vehicle classes, the system outperformed YOLOv5, YOLOv7, and YOLOv8 with a 96.4% detection accuracy at 125 FPS, which also decreased waiting time by 38 %. Detection accuracy nighttime for pedestrians boosted safety in low-visibility conditions to 92.5 %. The system also reduced <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{CO}_{2}$</tex> emissions by 35 % while maintaining 15 ms latency with 5 G transmission, achieving real-time decision making, reinforcing sustainability goals. The system outperformed other models in high-density traffic regions, and served as a proof of concept to show deep learning, IoT, and real-time optimization can improve urban mobility, congestion, and smart sustainable city planning.