Variably Trimmed Mean CFAR Detector for Nonhomogeneous Background. (Final Report, July 1988-May 1989).

摘要

This thesis investigates the performance of various constant false alarm rate (CFAR) detectors in both stationary noise conditions and nonideal conditions caused by multiple targets and nonuniform clutter. In particular, we introduce a new scheme for CFAR detection and called the Variably Trimmed Mean (VTM) CFAR detector. We show that this detector is a CFAR detector in exponentially distributed as well as in more general clutter distributions. A closed form analytical expression for the false alarm probability in exponentially distributed homogeneous clutter is given. We also show that in the exponentially distributed clutter situation, this detector can be designed to yield performance characteristics that are better than those of the Order Statistic and Trimmed Mean CFAR detectors. (rh)