Terrain images segmentation in infra-red spectrum for autonomous robot navigation

Due to a wide spectrum of applications, autonomous ground vehicle navigation based on visual information has rapidly developed in the past decade. These applications include intelligence transportation systems, military robots and mars/lunar robot rovers. However, most of the attention was focused on navigating the robot in day light environment using visual spectrum images captured from CCD cameras. In this paper we propose a terrain segmentation system designed for navigating at night, as well as day time, using images in infra-red (IR) range. For this purpose, texture features based on Speeded-Up Robust Features (SURF) are extracted from IR images and their likelihoods are estimated using machine learning algorithms. The features´ likelihood and their spatial positions are applied for image segmentation.