Inversion is an important process in remote sensing. In order to improve the stability and accuracy of inversion, in this article, we applied kernel forms of AMBRALS (Algorithm for Model Bidirectional Reflectance Anisotropies of the Land Surface) and PLS (Partial Least Square) regression technique to simplify a canopy reflectance model SAILH (Scattering by Arbitrarily Inclined Leaves, with Hotspot effect). PLS is a statistical method used for regression highly collinear variable data. Kernel-driven model is a semi-empirical model with linearity form of "kernels", and these kernels can be explained in physics. We generated 24 typical canopy cover scenes by combining the canopy parameters of SAILH model. For each scene, we used PLS regression to estimate the coefficients of our new model. The results suggest the new model is acceptable in stability and accuracy. Base on the new model, we defined sensitivity matrix to assess the correlations of directional observations data, which can help to choose appropriate directions when inversion.

Date Published: 8 August 2007
PDF: 11 pages
Proc. SPIE 6752, Geoinformatics 2007: Remotely Sensed Data and Information, 675214 (8 August 2007); doi: 10.1117/12.760659

Published in SPIE Proceedings Vol. 6752:
Geoinformatics 2007: Remotely Sensed Data and Information
Weimin Ju; Shuhe Zhao, Editor(s)