Nonhomogeneity detection and the multistage Wiener filter

This paper introduces the multistage Wiener filter for radar space-time adaptive processing, combined with the generalized inner-product as a preprocessor in nonhomogeneous environments. By using recorded data from the Multichannel Airborne Radar Measurement program, the performance of the multistage Wiener filter and sample matrix inversion are assessed both with and without the preprocessor. The constant false-alarm rate test statistic is computed for each range bin and the performance metric used in this analysis is the ratio of the target value to the root mean square value of the noise values. Both high and low sample-support environments are considered. The reduced-rank multistage Wiener filter is demonstrated to outperform full rank sample matrix inversion, even with the generalized inner-product preprocessor. Additionally, the multistage Wiener filter is shown to have its largest impact when used in conjunction with the preprocessor in the low sample-support environment. In this case, it nearly achieves the performance obtained by the full-rank and high sample-support case.