Precision of Vegetation Height Estimation Using the Dual-Baseline PolInSAR System and RVoG Model With Temporal Decorrelation

This is the authors' abstract. We don't add key points for this paper.

Estimating vegetation height from polarimetric interferometric synthetic aperture radar (PolInSAR) data using the Random Volume over Ground model has motivated several studies. Most of these propose estimators and apply them to real data to demonstrate their potential. In previous publications on the single-baseline system, we proposed a complementary approach, which consisted in analyzing the precision of estimations of vegetation height that can be expected depending on the considered model and on the available a priori knowledge. In this paper, we develop such an analysis for the case of a dual-baseline (DB) system. We consider the DB configuration with a PolInSAR set obtained with three PolSAR acquisitions, the extinction coefficient of the volume is assumed unknown, and the level of temporal decorrelation is assumed to be unknown. The observed high sensitivity of the vegetation height Cramer-Rao bound (CRB) with respect to the system parameters and the vegetation characteristics shows that the system optimization cannot guarantee 1-m precision for all vegetation heights, even for large estimation windows with N=2000 pixels. Nevertheless, an operating regime exists for which the vegetation height estimation precision is around 1 m for N=200 pixels. This regime is obtained for a pair of wavenumbers (0.06 and 0.25 m-1), for vegetation height ranging [20, 50] m, and for polarimetric contrast between the ground and the volume larger than 0.3. Furthermore, we investigate the performance of a maximum-likelihood estimator and compare this to the precision given by the CRB. For the examples considered, with N=200 pixels, we observed convergence issues of the estimator when the polarimetric contrast is smaller or equal to 0.3.