Performance evaluation of parametric Rao and GLRT detectors with KASSPER and Bistatic data

The parametric Rao and GLRT detectors, recently developed by exploiting a multichannel autoregressive (AR) model for the spatially and temporally colored disturbance, were shown to perform well with limited or even no range training data for the airborne radar configuration. In previous computer simulation studies of these parametric detectors, the disturbance was generated as a multichannel AR process. However, the disturbance signal in an airborne radar environment do not necessarily follow an exact multichannel AR model. In this paper, we evaluate the detection performance of the parametric Rao and GLRT detectors using more realistic datasets: the KASSPER 2002 dataset that includes many real-world effects such as heterogeneous terrains, antenna errors and leakage, and dense ground targets/discretes, etc., and the Bistatic dataset which contains range-dependent clutter due to bistatic geometry. Experimental results on both datasets show that the parametric detectors can provide good detection performance with limited or no range training in more realistic radar environments.