Variational assimilation of UAV multispectral data for identifying the thermophysical properties of objects using nnU-Net and a finite element model
A method for variational assimilation of UAV multispectral data is proposed to identify effective thermophysical properties of surface objects from a limited set of optical and infrared observations. The method combines radiometric processing, reconstruction of daily temperature dynamics, semantic segmentation by a modified nnU-Net, identification of the heat flux function, and a coefficient inverse heat conduction problem based on a finite-element model. The segmentation map is used as semantic prior information for selecting the direct model structure and parameter constraints. Experimental testing was performed for an engineering campus area of 11.4 ha. The results demonstrate the possibility of constructing spatial maps of density, volumetric heat capacity, thermal conductivity and thermal inertia for 17 semantic object classes.
Authors
- Yu. Yu. Gromov
- V. V. Vologin
- I. N. Ishchuk
- E. E. Nazarov
- A. A. Solodovnikov
Institutions
- Moscow Institute of Physics and Technology (RU)
Publication Details
- Journal
- INFORMACIONNYE TEHNOLOGII
- Published
- 2026-11-20
- DOI
- https://doi.org/10.17587/it.32.590-598
- Citations
- 1
- Primary Topic
- Urban Heat Island Mitigation
- Type
- article
- Field-Weighted Citation Impact
- 1.27