| Fully polarimetric synthetic aperture radar(POLSAR),as an advanced technique for earth remote sensing,pertinently enhances the probing of target’s geometry structures and dielectric properties.In particular,incoherent target decomposition is powerful to extract the scattering characteristics of the target from fully polarimetric SAR image.However,there is no rigorous practice,in theoretical sense,to measure the information content of the coherence matrix on which the target decomposition is performed.Also,the size of moving window for calculating the statistically averaged coherence matrix is heuristically determined.This study attempts to resolve the above-mentioned issues by applying the information entropy,commonly used in communication theory,inspired from the context that the core of remote sensing is essentially to maximize the target information content from a limited observations.We first examine the Shannon entropy,and then formulate it as a measure of information content in coherence matrix,while,by measuring the entropy,choosing the good size of moving window that is used to estimate the statistically averaged coherence matrix.To validate the effectiveness of the proposed approach,we test data sets from Cband Radarsat-2,C-band GF-3,and L-band AIRSAR,containing various land-covers.Results indicate that information content of the coherence matrix varies for different land-covers,as expected.For built up land,the information content is low,relatively,and increases linearly with increasing moving window size,and for ocean surface and vegetation canopy,information content is high,and,similar to those for building,increases more quickly with the window size,but tends to saturate thereafter.For the same land-cover,the tendency of the information content remains quite similar,regardless of the data sets in line of frequency bands and dates.In summary,Shannon entropy is a simple and yet effective measure for information content of the coherence matrix in the context of target decomposition,and accordingly the window size can be optimally selected.To this end,deeper physical insights into the coherence matrix can be better understood and,consequently,more potential applications of the POLSAR may be exploited. |
