Methods for Detecting and Eliminating Undesired Target Returns from Training Data in Adaptive Radar Processing

"Recent studies have shown that the performance of Space-Time Adaptive Processing (STAP) degrades when there are moving targets in the secondary training data. Although a variety of techniques have been proposed for controlling the effects of targets inthe secondary data (TSD), these techniques generally do not take advantage of external information. We are developing knowledge-based STAP (KBSTAP) approaches that include three techniques for predicting regions in the CPI data cube that are likely tocontain targets: (1) use ALPHATECH's MHT Tracker to project target locations from previous CPIs, (2) transform information about roads and traffic speed distribution into the CPI coordinate frame, and (3) use the output of STAP itself in an iterative loopto estimate and then remove targets responses. We are also developing two techniques for removing target responses from the training data: (1) apply a re-weighted covariance estimator after excision; (2) coherently suppress the target response using avariant of the CLEAN algorithm and estimate the clutter covariance after omitting the contaminated cells. In Phase II will refine the algorithms and demonstrate their effectiveness using high quality simulation as well as actual data."