The Research Of Cfar Detection Algorithm Based On Knowledge Aided Technology

CFAR (Constant False Alarm Rate) is a detection algorithm in radar targetdetection which provides an adaptive threshold, the radar system can maintain constantfalse alarm probability by using CFAR algorithm. In CFAR detection, In order to get theoptimal estimation of distribution characteristic of cell under test (CUT), CUT andreference cell should satisfy the conditions of independent and identically distributed(i.i.d.). However, the presence of nonhomogeneous and multi-distributions backgroundsin actual clutter environment destroyed the applicable conditions of a traditional CFARdetectors, will lead to degradations of detection performance and CFAR performance.Knowledge aided (KA) is a technology by mining the prior information of complexenvironment and target and using the prior information in signal processing, whichcould improve the performance. A large number of studies have shown that terrainfeature of resolution cell is the main factors to affect the clutter power distribution. Thus,the clutter distributions of the same or similar terrain areas can be considered arehomogeneous. Using terrain feature as knowledge aided in CFAR detection, could solvethe problems of nonhomogeneous clutter.This paper focuses on the research of knowledge aided CFAR detection, the majorwork of the paper mainly includes:1. The mathematical model of clutter and characteristics of the target in radarsignal detection is discussed, the background, significance and framework of traditionalCFAR algorithms is researched, above all become the foundation of the paper.2. Due to the target detection in nonhomogeneous gaussian clutter background, theprinciples and detection performances of Mean Level (ML) CFAR and OrderedStatistics (OS) CFAR are discussed. Knowledge aided technology could improve theperformance in nonhomogeneous clutter background. The validities of the KA-CFARalgorithm is proved by simulation.3. Due to the detection in nonhomogeneous non-gaussian clutter background,based on the research of clutter distribution characteristics about two-parameter modelsuch as Weibull and Log-normal clutter, the performances of ML, OS and BLUE CFAR in the two backgrounds are analyzed. On this basis, Knowledge aided technology couldimprove the parameter estimation accuracies and detection performances ofdual-parameters CFAR, the validities of the KA-CFAR algorithms are proved bysimulation.4. Due to the situation of clutter distributions are unknown in projects,Anderson-Darling goodness-of-fit test is analyzed. On this basis, computationalcomplexity of Anderson-Darling test is optimized and the solution of fitting conflict isdesigned. By using the knowledge aided, the method of distribution familyidentification for the clutter edge is designed. The validities and real-time of theKA-CFAR algorithms are proved by simulation.The effectiveness of the work has been verified by simulations. The results showthat the CFAR detection technique based on knowledge aided technique can both keepthe false rate constant and detect the target effectively.