Multiscale Forest Structure Estimation from SAR Tomography

Abstract:
Recently a framework for forest structure based on TomoSAR data has been defined. Based on the spatial distribution of the reflectivity peaks, two complementary descriptors are evaluated at a given scale. This paper highlights the dependence of the structure estimate on scale and shows, through analysis on real and simulated data that the restriction to one single scale does not always ensure a complete characterization of structure. As a consequence, a multiscale framework, accounting for the evolution of the estimated structure across scales is suggested for a better forest structure assessment. Furthermore, the effect of system resolution on the accuracy of the estimation is quantified and thus system requirements for forest applications are determined.