Forest Mapping and Classification at L-Band using Pol-InSAR Optimal Coherence Set Statistics

This paper presents a new approach to classify forested areas from POL-inSAR data. The statistics of an optimal coherence set are derived to define a log-likelihood distance that can be used in iterative classification processes. This novel method is compared to an existing technique based on a POL-inSAR coherency matrix Wishart statistics. The invariance properties of optimal coherences may be used to overcome some limitations encountered with theWishart approach. It is then shown that such an approach may be used to relyably classify forest stand biomass into broad categories.