Physic based aggregation model for the unmixing of temperature and optical properties in the infrared domain

The new generation of airborne hyperspectral image sensors sensitive in the thermal infrared domain gives the opportunity for new urban applications like classification, thermal budget at building level, etc. Nevertheless, due to the high spatial variability of such landscape, the signal acquired by each pixel remains complex as it results of the contribution of several radiative signals coming from different materials. The understanding of such signal requires the developement of an aggregation model that modelize the optical properties and surface temperature aggregating at different spatial scales. This paper proposes to justify the unicity of the aggregation model, proposed by Fontanilles et al, on LWIR spectral range (8 - 14 μm) for small temperature variations (≤ 20K).