Implementing cluster analysis to knowledge-aided covariance estimation

This paper proposes a knowledge-aided covariance estimation approach for space-time adaptive processing to deal with the performance degradation in heterogeneous clutter environment. In this approach, considering the fact that the clutter scattering power often varies considerably from different terrain types, but changes little from the similar terrains, an adaptive cluster analysis on clutter scattering power based on K-means algorithm is proposed for reasonable terrain classification in which an evaluating indicator is imported to cluster processing to acquire the optimal classification result. Then the scattering power of all clutter patches within each type of terrain is averaged and the averaged clutter power is transformed to clutter covariance matrix via the space-time steering vector estimated by priori knowledge. The simulation results assess the performance of this approach and confirm the validity of proposed adaptive cluster analysis and the effectiveness of covariance estimation.