Conventional and distributed CFAR detection in nonhomogeneous background

In this dissertation, we consider CFAR processing in different configurations and/or environments. We develop the theory of distributed CFAR detection with data fusion employing OS-CFAR processors as local detectors in a Gaussian environment. We find optimal values of the local detector parameters for a given fusion rule in a homogeneous background. Then, we compare its performance with the performance of single sensor OS-CFAR detection as well as with the performance of distributed CA-CFAR detection. For comparison purposes, we use the same configuration and assumptions as for our distributed OS-CFAR detection model both in homogeneous and nonhomogeneous backgrounds.; We introduce a new CFAR algorithm that is a generalization of the well known robust OS-CFAR algorithm. The new algorithm uses a linearly weighted sum of selected ordered statistics (SOS). It provides an unbiased minimum variance estimate of the noise power level of the background. We describe the operation of the SOS-CFAR and derive the exact analytical expressions for the resulting probability of detection and false alarm both in homogeneous and nonhomogeneous backgrounds. Numerical results are obtained to compare the performance of SOS-CFAR with the performance of OS-CFAR under the same conditions both in homogeneous and nonhomogeneous backgrounds.; We analyze the probability of false alarm performance of the CA-CFAR and OS-CFAR processors in an environment where clutter power varies linearly and the background is Gaussian. In numerical examples, several cases are considered such that the linearly variant part of the clutter power occupies less than, equal to, or more than the reference window of the CFAR processor.; We also consider adaptive CFAR detection in a K-distributed clutter-only environment. We introduce two new CFAR algorithms designed for this particular environment. The first algorithm uses the Mean Square estimate (MS-CFAR) and the second one uses the Maximum A Posteriori estimate (MAP-CFAR) of the mean level component of the K-distributed clutter to set the adaptive threshold of the CFAR processor. We derive the analytical expressions for the MS and MAP estimators and analyze their performances by Monte Carlo simulation. We also compare the performances of MS-CFAR and MAP-CFAR processors with that of the CA-CFAR processor.