Ground moving target detection for along-track interferometric SAR data

This paper investigates different metrics for ground moving target indication (GMTI) with a multichannel synthetic aperture radar (SAR). These metrics are especially sensitive to targets that have an across-track component of velocity, v_y, although they can also detect targets with an along track velocity component, v_x. Not only are the metrics suitable for fully SAR compressed data, they are also meaningful when computed for range (pulse) compressed but Doppler uncompressed data (raw data). The use of both data domains allows for target detection over a larger v_y range than either used individually. For large v_y, side-lobe suppression in fully SAR compressed data simultaneously suppresses the targets, giving an advantage to the raw data domain. For low v_y, and especially with low RCS targets, the SAR compression gain facilitates detection, giving an advantage to the SAR compressed domain. By viewing the SAR compression as the application of a linear filter, the metrics are shown to be suitable for target detection after any linear filtering of the raw data, such as a linear time-frequency filter, or a simple transformation into the Doppler domain using an FFT filter. The metrics have in common that they are all based on the eigenvector decomposition of the sample covariance matrix. Their statistical properties are analytically compared and their detection capabilities are demonstrated on measured two-channel airborne SAR data in a variety of data domains. Some metrics are shown to allow target detection with low false alarm rates even in heterogeneous terrain, such as in urban areas, without compromising the probability of detection.