Analysis Of Some Cfar Processors For High-resolution Radar Systems

Résumé: In this study, we address the problem of adaptive detection of embedded radar targets in a Pareto distributed clutter. This type of detection is achieved by maintaining a constant false positive rate (CFR) during processing. We first introduce the geometric mean CFAR (GM-CFAR) detector introduced in the literature. The detector is suitable for homogeneous clutter. We show that the derivation of this detector is obtained by exploiting the duality between both exponential and Pareto distributions. This duality makes it possible to convert CFAR detection strategies developed for Gaussian noise into Pareto distributed clutter. In addition, we have presented a modified version of the GM-CFAR, which is called the Maximum Likelihood (ML) CFAR detector, and thus give the corresponding analytical expression of the false alarm probability. For comparison purposes, we perform as well a partial theoretical analysis of two other detectors, namely the Greatest Of (GO) CFAR and the Smallest Of (SO) CFAR algorithms. Via numerical simulations, the performance of the ML-CFAR scheme has been compared and examined against those of the Optimal, GM, GO and SO-CFAR detectors. The simulation results obtained validate the interest of the ML-CAFR detector in homogenous backgrounds, but shows that its performance degrades in heterogenous backgrounds. Moreover, it has been also shown from these results that the SO-CFAR detector is relatively efficient against interfering targets.

Publié dans la revue: