Speed Sign Detection Using Computer Vision Techniques in ADAS Systems

Abstract

Traffic sign detection is one of the most important modules in Advanced Driver Assistance Systems (ADAS) that help to improve road-safety. In this paper, the real-time traffic sign detection system is proposed based on a Convolutional Neural Network (CNN) model implemented in Python. The system identifies and classifies traffic signs in video feeds while handling variations in object size. By convolutional layers, feature extraction was optimized to improve the detection of small and occluded signs. A special loss function is also used to enhance the accuracy of bounding box regression. Much testing on a real-life traffic sign dataset exhibits that the system reaches the best accuracy of 97%, which remarkably reduces detection under challenging conditions with a low miss rate and high overall accuracy.