Support Vector Machines for Road Images Recognition in Autonomous Car

This paper presents performance of Support Vector Machines (SVMs) for recognizing road images. All images are collected by our car. They are divided into 4 classes turn left, turn right, forward and stop. We use 90% of them for training and use 10% for testing a model. An accuracy of model is about 70.77%.

I. Introduction

In present, transportation is very convenient. There are many options for traveling. People can travel to anyplace in quite short time. In Thailand, most of people use the car for travel because of their convenience. As a result, the number of cars in Thailand is increasing. Moreover, the number of accident increases as well. According to traffic accident statistics Fig. 1:

Traffic accident statistic [1]