Research On The Techniques Of Radar Constant False Alarm Rate Detection Under Complex Clutter

CFAR (Constant False Alarm Rate), as an adaptive threshold detection algorithm designed for detection under complex clutter environment, aims at acquiring desired and predictable detection probability while maintaining constant false alarm probability.This paper focuses on the CFAR detection techniques under complex clutter background. Clutter samples from different platforms, including airborne Pulsed-Doppler (PD) clutter, IPIX sea clutter, and MSTAR SAR (Synthetic Aperture Radar) clutter, are investigated. Several theoretical analysis and implementation means like hypothesis test, edge detection and texture segmentation are employed to perform clutter tracking and distribution family identification with the hope to improve the performance of CFAR detector and extend its usage in harsh clutter environment where clutter edge or interfering targets are present.The major work of this thesis includes:(1) Basic radar CFAR detection theory is reviewed and the clutter environment is investigated and classified as homogenous and non-homogenous background. The performance of conventional Mean-Level(ML) CFAR and Order-Statistic(OS) CFAR detectors are analyzed and summarized in details, offering benchmarks for further CFAR detector design and performance evaluation.(2) Clutter tracking strategies for different clutter types are investigated. For airborne Pulsed-Doppler radar clutter, detailed characteristics of PD radar clutter regions are studied by the both simulated and real clutter data. A Gradient Vector Flow (GVF) Snake based clutter tracking algorithm is proposed to track the main-lobe and altitude line clutter regions in the grayscale clutter map. The relationship between the clutter boundary and airborne platform parameters are studied in detail to boost clutter tracking by initializing the Snake spline near the desired clutter boundary. For two dimension clutter data like IPIX sea clutter and MSTAR Synthetic Aperture Radar (SAR) data, the clutter tracking work is done by the multifractal transform and fuzzy cluster to the resultant multifractal exponents.(3) A clutter tracking based CFAR detector (CT-CFAR) is designed, by taking advantage of clutter tracking results to determine if clutter edges or interfering targets are present in the reference window. Performance of CT-CFAR is illustrated by detection simulations under both homogenous background and clutter edge scenarios. The relationship between CT-CFAR performance and precision of cluter tracking is also discussed.(4) Common clutter distribution families and parameter estimation methods are introduced and analyzed. Anderson-Darling goodness-of-fit test is used for clutter distribution family identification for Rayleigh, Weibull, Lognormal and K-distribution. The corresponding critical value table for the underlying families is composed and adjusted for distribution identification purpose. Derived from the Anderson-Darling distribution test, K-sample Anderson-Darling hypothesis test is used for homogeneity test between clutter regions. An Anderson-Darling test based CFAR detector (AD-CFAR) is designed by the combination of these two means, which performs K-sample Anderson-Darling homogeneity test in the reference window to search and merge small homogeneous clutter blocks to a larger one and proceeds to Anderson-Darling distribution identification. Thus the final detection strategy is selected from the CFAR strategy library according to the identification results. The designed detector has the advantage of low CFAR loss in homogenous background, low false alarm peak at clutter edge and robust performance when interfering targets are present.A brief conclusion is made at the end of this thesis; the outlook of CFAR detection under complex clutter and issues that need further investigations are also covered.