Real-Time Traffic Sign Recognition For Smart Transportation In Indian Urban Environments Using Yolov9

Abstract

Traffic sign detection and recognition are the two main factors in the development of intelligent
transportation systems and autonomous vehicles. This paper introduces a novel approach for real-time
traffic sign detection and recognition, utilizing the You Only Look Once version 9 (YOLOv9) algorithm for
enhanced efficiency and accuracy. YOLOv9 is an advanced deep learning model that contributes significant
improvement in detection accuracy and processing speed compared with its predecessors, including
YOLOv5, YOLOv7, and YOLOv8. The proposed model is specifically designed to address the diverse and
complex traffic environment encountered in Indian cities, where traffic signs are frequently nonstandardized,
inconsistently placed, and locally modified. By leveraging a robust dataset of Indian traffic
signs — collected from urban, suburban, and rural areas across the country — the model detects and
classifies multiple sign types with high precision. Data augmentation strategies including rotation,
brightness adjustment, occlusion simulation, and noise addition further strengthen the model's robustness
against challenging real-world conditions. The system integrates with smart traffic management
infrastructure, enabling dynamic control of traffic signals and supporting Advanced Driver Assistance
Systems (ADAS). Experimental results confirm that the proposed YOLOv9-based method achieves higher
detection accuracy, superior mean average precision (mAP), and faster processing speed compared to
existing models such as YOLOv5, YOLOv7, YOLOv8, SSD, and Faster R-CNN. The proposed approach is
therefore well-suited for practical deployment in Indian urban transportation systems.