Image Feature Based Machine Learning Approach for Road Terrain Classification

The type of road terrain is an important piece of environment perception information for smart vehicles. In this paper, an approach to recognize the road terrain by using road surface images is proposed. The texture features were extracted from images captured by a camera for the road terrain classification task. The support vector machine method was selected for terrain classification. The simulation experiments indicated that this approach worked very well on terrains with distinguished texture. However, this approach could not recognize the road terrain when the damaged captured image changed frequently due to the luminance or when the moving speed of the vehicle was too fast.