Investigation on Forest Height Extraction Using the Polarimetric Interferometric SAR Images

The work aimed in this paper investigates the applicability of forest parameters retrieval using the polarimetric interferometric synthetic aperture radar (PolInSAR) techniques. The application of PolInSAR in the forest area is based on the separation of backscattering phase centers associated to the top of canopy and the terrain. In this regards, one important and widely used technique is the random volume over ground model in the two-layer forest scenario. The unknown parameters of the model such as terrain topography, forest height, extinction, and effective ground to volume ratio, should be solved through a full nonlinear algorithm. However, due to the nonlinear nature of the problem and its complexity as well, the obtained solution depends strongly on the choice of the starting value and the considered polarizations. In the literature, to simplify the problem some important assumptions have been considered. However, the result is also highly affected by the considered assumptions. To address these problems, a simple model-based strategy is presented to obtain the forest parameters in the well-known random volume over ground model. The method obtains the solution without any assumption regarding the parameters. The experimental results show that the framework provides a promising solution for forest parameters extraction in comparison with the inversion and SPIRIT methods. The root mean square error of forest height estimation is decreased from 13.26 and 11.41 in the inversion and SPIRIT methods, respectively, to 7.34 meters in the proposed model.