Robustness of polarimetric detection algorithms against steering mismatches

We analyze robustness of several multichannel radar detection algorithms to mismatches in the steering vector. In particular, we address the scenario of a staged detection, where a history of measurements provides the region of interest and a-priori polarimetric target signatures for a target. The four second stage detection approaches - optimally weighed span, Kellypsilas and Robeypsilas detectors, and Gerlachpsilas secondary data free detector - are considered using these polarimetric features. The evaluation based on recordings of real natural scenes and both real and artificially inserted extended objects concerns robustness of these algorithms to mismatches in the steering vector. It has been observed - for different ground clutter environments into which extended target is placed - that Kellypsilas detector provides the best detection but may suffer severely from the mismatch, Gerlachpsilas detector while allowing higher probability of false alarms, produces the most stable performance and Robeypsilas matched filter gives a suitable trade-off.