Texture Estimation Using Thermography and Machine Learning

Abstract

Contactless material characterization has the potential to be used in various applications such as teleoperation and autonomous physical interaction robotics. Active infrared thermography is a promising approach for classifying materials based on their thermal response to laser excitation over a short distance, thus creating a contactless haptic modeling scheme. However, factors such as the texture of the object under inspection can influence the thermal signature and therefore need to be compensated against. This paper presents a method to use the exact components of a thermographic material characterization system to estimate texture, allowing it to produce more robust characterization in the presence of textured surface. Experimental results confirm that the system is capable of estimating the texture of the sampled material surface to a sufficient degree, with a promising outlook for further improvements as the data set is scaled.