Preiss, Mark; Stacy, Nick J. S. (Defence Science and Technology Organisation, PO Box 1500 Edinburgh 5111, Australia)
In repeat pass interferometric SAR man-made scene changes may be detected by identifying regions of low coherence. The ability to detect such changes however, depends on other sources of decorrelation such as environment effects and baseline decorrelation. Recently a Log Likelihood Change Statistic (LLCS) has been derived by formulating the problem in a Bayesian hypothesis testing framework. In this paper the LLCS is extended to exploit the additional information available in polarimetric SAR imagery. It is shown through simulation and application to experimental data that the fully polarimetric LLCS significantly improves the ability to detect scene changes.