A Comparison of Methods to Detect and Eliminate Secondary Target Returns from Adaptive Radar Training Data

For SAR, spatial and temporal adaptive processing (STAP) uses samples local to the cell being processed to estimate local clutter and noise statistics. If these samples are corrupted by returns from other unknown targets, the statistics are invalid and theperformance of the algorithms can be degraded. We propose to conduct a range of experiments to evaluate signal detection, statistical, multi-look, and tracking approaches for detecting unknown target returns in the training data. These include the CLEANalgorithm for detecting and eliminating returns by iteration, a statistical jackknifing approach to detect anomalous training data cells, combining scans from different angles to optimize detection, and near-clutter tracking methods such as