Band selection for mid-wave infrared thermography through extended radiometric calibration and heuristic-based optimization

Non-contact radiation-based multispectral thermometry methods are inherently underdetermined due to the inability to decouple the effect of emissivity and temperature on the spectral exitance of the target. This work introduces a calibration method that incorporates the assumption of local emissivity linearity, leading to an extended radiometric system model with better regularization of the inverse problem. A heuristic-based multi-objective optimization scheme for designing optimal filter configurations per target scenario is developed, which considers the spectral shape of the band-pass filters. Through a per-scenario suggestion of filter combinations, the suggested procedure can use prior information about the spectral emissivity and the temperature range of interest. The optimization procedure is demonstrated for a four-channel thermographic system and numerically evaluated over a broad range of common aerospace materials. An optimally designed system, operated according to the suggested guidelines, is predicted to result in temperature recovery errors within 10 K over a range of 423–873 K.