Performance Evaluation of Different Machine Learning Techniques with Stereo Vision Used to Road Detection Task

Road recognition using visual information is an important capability for autonomous navigation in urban environments. Over the last three decades, a large number of visual road recognition approaches have been appeared in the literature. This paper proposes the use of depth information obtained from stereo camera along with other features based on color to detect road. Several features were evaluated using selection methods and its derived subsets were used to test some machine learning techniques in classification task. We used averages, entropy and energy from RGB, HSV, YCbCr and normalized RGB and mean of disparity as input features. Experimental tests have been performed in several situations in order to validate the proposed approach.