AI-Driven Speed Limit Sign Recognition and Real-Time Adaptive Speed Limiting System

Abstract

Road safety continues to be affected by drivers missing or misinterpreting roadside speed limit signs, particularly under challenging traffic and environmental conditions. This work presents an intelligent speed regulation framework that combines vision-based traffic sign recognition with automatic speed control. A YOLOv5m object detection model is employed to identify traffic signs from video frames, while Optical Character Recognition (OCR) extracts the numerical value from detected speed limit signs. The recognized speed information is transmitted to an embedded controller that dynamically adjusts the vehicle’s operating speed. To evaluate the proposed approach, multiple detection models were examined, with YOLOv5m providing the most favorable balance between recognition accuracy and execution speed for real-time deployment. The framework was trained using an augmented traffic sign dataset containing Indian road signs together with additional custom images. Experimental evaluation demonstrated a precision of 93.3%, recall of 90.2%, and mAP@0.5 of 95.5%, indicating reliable performance across diverse traffic scenarios. The integration of computer vision and embedded control demonstrates a practical and economical solution that can contribute to safer driving and future intelligent transportation systems.