Методи комп’ютерного зору для розпізнавання дорожніх знаків у системах інтелектуального транспорту

Abstract

Traffic sign recognition is an important task in modern intelligent transportation systems and autonomous vehicles. Accurate detection and classification of road signs allow vehicles to correctly interpret traffic rules and improve road safety. This paper considers computer vision methods used for solving the problem of traffic sign recognition. The general pipeline of such systems includes image acquisition, preprocessing, sign detection, feature extraction and classification. Traditional approaches based on color and shape analysis are widely used for detecting candidate regions of road signs. Feature extraction methods allow systems to describe visual characteristics of detected objects. Machine learning algorithms are then applied for classification of traffic signs. Recently, deep learning approaches, especially convolutional neural networks, have significantly improved recognition accuracy. These methods allow systems to work reliably even in complex real-world conditions