Road Segmentation Based on Learning Classification

In vision navigation tasks, the road segmentation is a useful method. Usually, roads can be detected using image segmentation and related image process methods. However such methods always rely on specific road prior knowledge, and they are difficult to be realized in different environments. In this paper the learning classification is proposed. In the learning process the roads of different environments are labeled and learned. In the classification process the road is segmented and selected. Experiments results show that even the roads change the correct results could be got for online process.