Alternatives to the Normalized Matched Filter in Multitarget Scenarios based on Robust Noise Estimators

The Normalized Matched Filter (NMF) is widely used in radar for its Constant False Alarm Rate (CFAR) property with respect to unknown noise power. While optimal for a single target, we show that its performance degrades in multitarget scenarios, even when targets are well separated, due to the coupling of the detection statistic with the total signal energy. This paper proposes a class of robust detectors that replaces the standard noise power estimator with alternatives based on spectral-domain robust statistics, such as quantiles and trimmed means. The resulting detectors maintain the CFAR property while improving detection performance in multitarget contexts.