An Optimal Nonnegative Eigenvalue Decomposition for the Freeman and Durden Three-Component Scattering Model

作者：Lim Yu Xian^*; Burgin Mariko S; van Zyl Jakob J

来源：IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(4): 2167-2176.

DOI：10.1109/TGRS.2016.2637882

• Model-based decomposition
• optimal nonnegative eigenvalue decomposition (NNED)
• radar polarimetry

Model-based decomposition allows the physical interpretation of polarimetric radar scattering in terms of various scattering mechanisms. A three-component decomposition proposed by Freeman and Durden has been popular, though significant shortcomings have been identified. In particular, it can result in negative eigenvalues for the component terms and the remainder matrix, hence violating fundamental requirements for physically meaningful decompositions. In addition, since the algorithm solves for the canopy term first, the contribution of the canopy is often over-estimated. In this paper, we show how to determine the parameters for the Freeman-Durden model in a way that minimizes the total power in the remainder matrix without favoring any individual component in the model, while simultaneously satisfying the constraints of non-negative eigenvalues. We illustrate our analytical solution by comparison with the Freeman-Durden algorithm, as well as the nonnegative eigenvalue decomposition (NNED) proposed by van Zyl et al. The results show that this optimum algorithm generally assigns less power to the volume scattering than either the original Freeman-Durden or the NNED algorithms.

• 出版日期2017-4