Temperature Correction and Reflection Removal in Thermal Images using 3D Temperature Mapping

Abstract

Many mobile robots nowadays use thermal imaging cameras (TICs) in order to enhance the environment model that is created during exploration tasks. In conventional thermography, thermal images always have to be carefully revised by human operators, which is not practicable in autonomous applications. Unknown surface emissivities are the main source of misinterpretations in thermal images. In this work, we present two methods dealing with these misinterpretations by exploiting the TIC’s changing point of view. While the first approach classifies the regarded material in order to estimate improved surface temperature values, the second one is capable of detecting and removing thermal reflections. The spatial relationship between the thermal images and the regarded surface is made by using a rigidly mounted sensor stack consisting of a TIC and a 3D laser range finder, whose extrinsic calibration is described. During evaluation, we demonstrate the functionality of both approaches.