Resumen

PAZ PELLAT, Fernando. Relative atmospheric corrections of satellite images: invariant patterns and atmospheric models. Terra Latinoam [online]. 2018, vol.36, n.1, pp.1-12. ISSN 2395-8030.  https://doi.org/10.28940/terra.v36i1.228.

To use information obtained with satellite technology reliably, it is necessary to eliminate or reduce the disruptive effects associated with the spectral information captured by sensors on space platforms. In this paper we analyze the inversion of radiative models of the atmosphere, which consists in determining the additive and multiplicative constants in each spectral band to make the necessary atmospheric corrections. The methodology proposes the use of invariant patterns of soil lines and dense vegetation for the inversion of radiative models. The results showed that, without knowledge of the atmospheric model or the type of aerosol, soil line data were relatively insufficient (low correlation) to obtain the additive and multiplicative constants of the atmospheric inversions, with problems of multiple solutions in the inversion process. Under similar conditions, the same was found for additive constants with the dense vegetation line, but for the multiplicative constants the results were favorable (R2 > 0.9). In contrast, with the knowledge of the atmospheric model and the aerosol model, estimates of additive and multiplicative constants were highly satisfactory (R2 > 0.99) in both cases. For soil line inversions, only one constraint of the two available was used. In conclusion, the use of invariant soil-line patterns allows us to establish two basic relationships to invert the radiative simulations of the atmosphere, prior to functional compaction, and field measurements can be made so that the proposed atmospheric correction process in this work can be considered in absolute and not relative terms.

Palabras llave : remote sensing; spectral patterns; soil lines; dense vegetation lines; simple restrictions.

        · resumen en Español     · texto en Español     · Español ( pdf )