Affine transform representation for reducing calibration cost on absorption-based LWIR depth sensing

Abstract

Multispectral long-wave infrared (LWIR) ranging is a technique that estimates the distance to the object based on wavelength-dependent absorption of LWIR light through the air. Prior works require time-consuming measurements for calibration and solve non-linear inverse problems, which sometimes falls into a local minimum. In this paper, we propose a linear representation that connects the measurements and the scene parameters using the affine matrix. In this representation, the distance and the temperature of the object can be obtained as a closed-form solution and the calibration cost can be reduced to at least three observations. In real-world experiments, we demonstrate that our method is effective to reduce the calibration cost while keeping the precision of the depth estimation.