Modified SDD-GLRT detector for range-spread targets in non-gaussian clutter

Scatterer density dependent generalized likelihood ratio test (SDD-GLRT) performans well in detecting a spatially distributed target in the presence of additive non-Gaussian clutter, where the clutter is modeled as a spherically invariant random vector (SIRV). But SDD-GLRT is a statistically optimum detector, which results in performance loss in the practical case. Base on the order statistics detection theory, a modified SDD-GLRT detector is proposed, which utilizes a priori knowledge of the scatterer spatial distribution fully. It shows that the probability of false alarm is independent of the underlying mixing distribution, the normalized clutter covariance matrix and the steering vector. Hence the modified SDD-GLRT is a constant false alarm rate (CFAR) detector. The experimental results show that the modified SDD-GLRT detection performance is significantly improved by increasing the number of scatterers or the number of pulses used. For detecting a sparsely distributed target, the modified SDD-GLRT outperforms both of the SDD-GLRT and the NSDD-GLRT. Furthermore, the modified SDD-GLRT is more robust for a moderate mismatch of the expected number of scatterers or different spikiness of the clutter.