Robust Polarimetric Adaptive Detector against Target Steering Matrix Mismatch

A polarimetric scheme is presented for the detection of a range-spread target when the presumed target steering matrix is deviated from the actual one. Following the adaptive beamformer orthogonal rejection test model, the decision statistic of the proposed detector involves the polarimetric clutter-plus-noise covariance matrix, the target component after noise whitening and rotation, as well as the sidelobe interference vectors of multiple range cells. In the practical realization of the detector, the polarimetric clutter-plus-noise covariance matrix is estimated by using the minimum mean-square-error regularized secondary data sample averaging. Taking into account the mismatch involved in the presumed target steering matrix, the target component is obtained by least square fitting subject to multiple separate conic uncertainty set constraints. The estimation of the sidelobe interference vectors is also based on the Bayesian statistics, and the involved polarimetric clutter-plus-noise covariance matrix is ultimately replaced by its Bayesian estimate previously obtained. The performance of the proposed detector has been evaluated and compared with the existing popular polarimetric detectors using both synthetic and real data. The results show that in the presence of target steering matrix mismatch caused by pointing errors, sensor position errors, channel calibration errors, multipath propagation, and random errors, the proposed detector has a superior performance to the current detectors in terms of the probability of detection.