Adaptive Radar Detection Algorithm Based on an Autoregressive GARCH-2D Clutter Model

We propose a model for radar clutter that combines an autoregressive (AR) process with a two-dimensional generalized autoregressive conditional heteroscedastic (GARCH-2D) process. Based on this model, we derive an adaptive detection test, called AR-GARCH-2D detector, for a target with known Doppler frequency and unknown complex amplitude. Using real radar data, we evaluate its performance for different model orders, and we use a model selection criteria to choose the best fit to the data. The resulting detector is not the constant false alarm rate (CFAR) with respect to the process coefficients, but we show that in practical situations it is very robust. Finally, we compare the AR-GARCH-2D detector performance with the performance of the generalized likelihood ratio test (GLRT), the adaptive linear-quadratic (ALQ), and the autoregressive generalized likelihood ratio (ARGLR) detectors by processing the real radar data. We show that the proposed detector offers a higher probability of detection than the other tests, for a given probability of false alarm.