Thermal Imaging for Robotic Applications in Outdoor Scenes

Abstract

This report describes thermal imaging techniques to identify different types of materials in outdoor scenes. First, we present the model that we developed to calculate the temperature of materials. By applying this model, we show how to calculate thermal inertia, a key physical characteristic of the material. Second, we analyze how applying vision segmentation algorithms to thermal images allows for discriminating between materials such as rack and sand. We then show how combining thermal and range information improves discrimination between occluding objects that are composed of the same material. Third, we examine how an autonomous robot can use these techniques to explore other planets. In particular, we show how a legged robot can use thermal inertia to select where to place its foot next or which material to sample.