Thematic classification with multitemporal thermal imagery

This work deals with thematic classification based on thermal data. We combine the theory of heat conduction of the Earth´s surface with the mathematical concepts of thematic classification. We introduce the ´Multitemporal Temperature Feature Space´ (MTFS), which is spanned by a set of temperature values measured at different times a day. The experimental evaluation is based on a measurement campaign that took place in the city of Nuremberg on 27 August, 1997. Airborne multispectral imagery was acquired with a DAEDALUS sensor at three different times: 4:30, 8:30 and 11:30 a.m. (MET). Classification results show a general applicability of techniques based on thermal data. For the classification of roofs and vegetation areas the MTFS is better suited than the reflectance space