Learning multi-static contextual target signatures

Target classification methods using multi-static radars have conventionally focused on the adaptation of techniques from mono-static active radar and ISAR techniques to the multi-static passive and active setups. One major facet of learning target classification using multi-static radar has received comparatively little attention: context. We propose to learn the contextual relevance of the diversity in the available data in multi-static setups. We test our approach in a passive multi-static setup. Our method first encodes relevant target features at single receiver sites and combines them to build a contextual signature of the target. We propose to harness the variation in SNR due to the bi-static RCS of the moving target for learning the target contextual signature. Our analysis suggests that robust classification of targets using such contextual signatures is possible and higher classification accuracies are achieved with increasing number of available sensors. We also benchmark our approach using geometric aircraft models.