The main objectives of this dissertation are to develop a constant false alarm rate (CFAR) detector which has excellent detection performance while keeping a desired false alarm rate in nonhomogeneous backgrounds and which requires less processing time than the other DFAR detectors based on order statistics, to study the performance characteristics of the proposed CFAR detector, and to develop the systolic array architecture of the CFAR detectors based on order statistics for real-time processing.
First, a modified order statistics (OS) CFAR detector called the order statistics cell averaging (OSCA) CFAR detector is proposed and analyzed for a Rayleigh target in nonhomogeneous backgrounds based on single-pulse detection. Then, we extend the analyses of the other modified OS CFAR detectors called the order statistics greatest of (OSGO) CFAR and the order statistics smallest of (OSSO) CFAR detectors for a RAyleigh target to nonhomogeneous backgrounds based on single-pulse detection. The computational complexity of each CFAR detector is also obtained in terms of the number of comparisons for ordering the samples in reference cells according to their magnitude. It is found that the OSCA CFAR detector performs better than the OS CFAR detector while the computational complexity of the OSCA CFAR detector is less than half that of the OS CFAR detector. However, the analysis results show that the OSGO CFAR and the OSSO CFAR detectors perform worse than the OS CFAR detector.
Second, we extend the analysis of the first part to multiple-pulse detection for chi-square targets. We study the performance characteristics of the OSCA CFAR and the OSGO CFAR detectors employing M-pulse noncoherent integration for chi-square targets. Explicit formulas for the false alarm and the detection probabilities for the OSCA CFAR and the OSGO CFAR detectors are also given. Analysis results show that the OSCA CFAR detector performs the best among the OS CFAR and the modified OS CFAR det...