Thai traffic sign classification system

Abstract

Nowadays, driving assistance technology is continuously developed to serve a comfortable and safe driving experience for the driver. The traffic sign is an important feature to improve the ability of this technology. However, in general, the traffic sign has various structures and details in each specific country for a clearly understanding purpose. Thus, this thesis aims to propose the classification and recognition system for the Thai traffic sign. In this study, the classification process provides the ability to categorize the Thai traffic sign into four classes by using a Support Vector Machine (SVM) and Random Forest (RF) classifier with the combination of two descriptor features that are the Histogram of Oriented Gradients (HOG) and Color Layout Descriptor (CLD) feature. In this proposed classification technique, it is able to reach the accuracy up to 93.98%. Besides, this study also presents the recognition process to recognize the type of each traffic sign. This process using two main techniques: Optical Character Recognition (OCR) and Normalized Correlation Coefficient (NCCoef) template matching to predict the real meaning of each sign in their class after classifying them.