Constant false alarm rate detection techniques based on empirical distribution function statistics
Constant False Alarm Rate (CFAR) techniques provide an adaptive threshold to distinguish targets from clutter interference in radar detection based on clutter statistics from a fixed-size reference window surrounding the cell under test (CUT). In this dissertation, empirical distribution function (EDF) statistics are used in goodness-of-fit tests to measure the homogeneity of the clutter sample in the reference window. New CFAR, techniques utilizing EDF statistics are presented which provide adaptive capabilities based on this measure, and improve average detector performance. These techniques vary the size of the reference window, or switch between alternate CFAR schemes in a composite architecture, based on EDF statistic tests. An adaptive reference window size provides more reliable clutter power estimates with large windows in regions of locally stationary clutter, and small windows in nonstationary regions. A composite CFAR architecture utilizes asymptotically-optimum Cell-Averaging (CA) CFAR in homogeneous regions or robust Ordered-Statistic (OS) CFAR in inhomogeneous regions. Multi-scale goodness-of-fit tests based on combined or multi-sample EDF statistics can provide information about clutter homogeneity as a function of scale, and provide the CFAR processor pre-sorted reference window samples at various scales. A modified OS-CFAR processor is presented which includes the CUT in the reference window to utilize pre-sorted samples as it moves between scales in a computationally efficient manner. Results indicate that this new technique performs well in nonstationary clutter environments of varying size, with a slight detection loss, and points to the feasibility of computationally efficient multi-scale OS-CFAR.
Bell, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our