On adaptive and distributed CFAR detection with data fusion

In radar signal detection, the problem is to automatically detect a target in a nonstationary clutter background while maintaining a constant false alarm rate (CFAR). One way to achieve CFAR is to adaptively set the detection threshold by estimating the noise level in the cell under test from nearby resolution cells. The presence of nonhomogeneities in the reference cells yield a poor estimation of the noise level. As a result, the probability of detection may be seriously degraded or an excessive number of false alarms may occur.; In this dissertation, new adaptive thresholding techniques for CFAR processing in nonhomogeneous background environments are proposed and analyzed. We develop various signal processing algorithms in which the samples in the reference window of the cell under test that may yield in a poor estimate of the noise level are effectively censored.; First, we evaluate the detection performance of the Censored Mean Level Detector, CMLD, for Swerling IV targets. The CMLD censors a fixed number of the highest ordered samples. Then, we propose the Generalized CMLD, where a censoring scheme determines and censors the interfering targets in the reference window. For multiple target situations in nonhomogeneous environment, the Generalized Two-Level CMLD, GTL-CMLD, is proposed next. The location of the clutter power transition is first determined, and then any possible interfering targets are censored.; For pulse-to-pulse partially correlated targets in nonhomogeneous clutter the Distributed GTL-CMLD is considered. Preliminary decisions made by individual GTL-CMLDs are combined at the data fusion center according to some "L out of M" data fusion rule.; The theory of cell averaging CFAR detection in spatially correlated and identically distributed clutter is then developed. We propose the Generalized CA-CFAR detector in which the threshold is set based on an estimate of the clutter covariance matrix and an estimate of the clutter level. Next, a novel censoring technique is considered for multiple target situations and spatially correlated clutter. Two preliminary decisions about each reference range cell are obtained by performing successive consistency tests on the degree of correlation between all adjacent pairs of reference range cells. For each range cell, these decisions are processed by the "AND" or the "OR" fusion rules in order to determine whether it contains interference or not.