Non-linear eigen-projection for SAR-GMTI in heterogeneous terrain

One particularly challenging area of implementation of Synthetic Aperture Radar-Ground Moving Target Indication (SAR-GMTI) is heterogeneous terrain. In areas such as urban terrain, an abundance of bright stationary targets greatly lengthens the tails of the clutter probability distributions for any metrics that are influenced by the power of the reflected radar return. This, in turn, raises the threshold for a given false alarm rate, thus lowering the probability of detection of the moving targets. In this paper, a new non-linear detector based on projection of the eigenvalues of the sample covariance matrix is shown to retain good probability of moving target detection in urban areas. Further, the new detector naturally polarizes bright stationary targets and moving targets to opposite sides of its domain. The statistics of the new detector, and examples of its use in urban terrain, are presented. The capability of the new detector to detect moving targets with a constant false alarm rate is compared to the classical method of the displaced phase centre antenna (DPCA) and a new metric called the hyperbolic detector.