Speaker
Description
Forest SAR tomography is a key tool for monitoring biomass. It involves processing a stack of coherent SAR images to extract various observables related to biomass. One challenge of forest SAR tomography is that the shape of the forest’s reflectivity profile is unknown. Typically, a prior is introduced; for example, the reflectivity profile is assumed to be Gaussian or an exponential. However, the choice of this prior affects estimation performance, because an incorrect model introduces estimation biases.
This communication focuses on characterizing both the forest and the ground backscattered responses using a recently introduced estimation method, called MoMET. The method is based on a characterization of the unknown reflectivity profile by its central moments. It is robust to uncertainty in the reflectivity profile, and is numerically efficient, allowing its application to large areas. It allows the estimation of the total backscattered power and the center of phase of the ground response, and the total backscattered power, the center of phase, and the spread of the forest response. The algorithm is applied to real tomography SAR data.
First, the use of central moments, which is the basis of the MoMET algorithm, is justified by a short discussion on the observability of the forest SAR tomography problem. The key idea is that the spread of the scatterers is small with respect to the system ambiguity, typically 5 meters vs. 90 meters. Therefore, the spectrum of the reflectivity density is smooth and can be approximated by the first terms of its Taylor expansion, which involves the moments of the distribution. The MoMET estimation algorithm is then presented, with a discussion on the choice of its hyperparameter, the maximal order estimated.
\medbreak
The algorithm is then applied to: $(i)$ TropiSAR campaign data and $(ii)$ Biomass mission data. To assess the performances, tomograms are computed, and MoMET estimates are compared to the LIDAR data available for the TropiSAR campaign and with the usual $L2$ parametric fits for both data sets. The results suggest that MoMET presents better robustness, better stability, and better numerical complexity than the usual parametric models.
Thanks to its robustness and its low complexity, MoMET algorithm is particularly suited for forest SAR tomography, where it can efficiently process wide areas.