Adaptive Cfar Detection of Targets in Non-Gaussian Clutter

Adaptive detection of radar targets in an unknown clutter environment relies exclusively on the classification of a statistical clutter model. Methods currently used for clutter model classification utilise maxi" likelihood based techniques, requiring large data sets. However since the data is generally non-stationary, only short segments of the received radar signal can be used in practice. There is a need for a robust classification strategy that is able to accurately disaiminate between clutter models when only short data segments are available. In this paper we present a constant fake alarm rate (CFAR) detection scheme which improves the power for classification of a clutter model under given situations. The proposed scheme utilises a knowledge-based approach which mask decisions made by a classifier based on the Kullback-Leibler mean information criterion for statistical model identification.