Recognizing articulated objects in SAR images

摘要

This paper presents the first sucessful approach for recognizing articulated vehicles in real synthetic aperture radar (SAR) images. This approach is based on invariant properties of the objects. Using SAR scattering center locations and magnitudes as features, the invariance of these features with articulation (e.g. turret rotation of a tank) is shown for XPATCH-generated synthetic SAR signatures and actual signatures from the MSTAR (public) data. Although related to geometric hashing, our recognition approach is specifically designed for SAR, taking into account the great azimuthal variation and moderate articulation invariance of SAR signatures. We present a basic recognition system for the XPATCH data, using scatterer relative locations, and an improved recognition system, using scatterer locations and magnitudes, that achieves excellent results with the more limited articulation invariance encountered with the real SAR targets in the MSTAR data. The articulation invariant properties of the objects are used to characterize recognition system performance in terms of probability of correct identification as a function of percent invariance with articulation.