Heterogeneity detection for hybrid STAP algorithm

Traditional STAP detectors require a secondary training data set that is target free and homogeneous with the cell under test (CUT). Hybrid detectors have been proposed for heterogeneous environments where the secondary data suffers from a statistical mismatch with respect to the interference in the CUT. These algorithms employ the generalised inner product (GIP) as a heterogeneity measure and eliminate the training data snapshots that are deemed heterogeneous. The GIP, however, does not take the presence of discretes or targets in the secondary data into account. If a target, or discrete, is present but sufficiently displaced from the signal of interest in the angle-Doppler plane, then it will not lead to any significant losses in the detection. Its presence, however, biases the GIP and leads to an undesirable rejection of the training data snapshot. This problem is examined in this paper where we propose the use of a projection-based statistic for heterogeneity detection. We show that this addresses the high target density problem.