Target detection in correlated SAR clutter

Synthetic aperture radar (SAR) clutter may be characterised by its single-point statistics and its correlation structure, both of which should be exploited in applications such as target detection. The non-Gaussian distributions associated with SAR clutter can be incorporated in detection schemes based on single-point statistics in a reasonably straightforward manner. However, use of the correlation structure is hampered by the difficulties associated with analytical manipulation of correlated, non-Gaussian distributions. In the paper, a generic model for correlated radar clutter is proposed based on a correlated Gaussian mixture distribution approximation to the clutter statistics in the log domain. Analytical manipulation of the model is possible because it is based on weighted sums of multivariate Gaussian distributions which have a tractable closed-form representation. The improved target detection performance which results from this approach is demonstrated.