Synthetic aperture radar (SAR) is a high-resolution radar imaging technique that is popular for many applications such as target recognition. The task of automatic target recognition (ATR) is difficult in general, but particularly in the case of SAR due to the high degree of image variability. This variability is due in large part to unmodeled aspect dependent reflectivities exhibited by scatterers on the target. Physical targets are composed of many of different fundamental scattering types, such as plates, corners, and cylinders, each of which has a specific aspect dependent reflectivity which varies with the scatterer shape, size, and orientation.
Our research addresses the issue of aspect dependent scattering for the purpose of ATR. Although this dependence in SAR data will limit performance when ignored or treated as a nuisance (which is common practice), we contend that it can actually be used to enhance ATR performance. Our attention particularly focuses on unimodal anisotropic scattering which is predicted by atomic scattering models for many common objects. In particular, we obtain a general characterization conveying the concentration and directivity of anisotropy. Not only does this characterization allow us to obtain more accurate estimates of the underlying reflectivity (due to the better reflectivity model), but the anisotropy classification itself can be used as a feature in target recognition. Degree of anisotropy is directly related to scatterer geometry for many scattering types, thus this attribution conveys novel information about the scatterer in question. Furthermore, the anisotropy attribution can assist the process of associating multiple pixels with a single physical scatterer as they will exhibit similar aspect dependencies.
A. J. Kim, and "Detection and Analysis of Anisotropic Scattering in SAR Data", Submitted to the Multidimensional Systems and Signal Processing's Special Issue on Radar Signal Processing and its Applications , Submitted November 2000.
A. J. Kim, and "Attributing Scatterer Anisotropy for Model Based ATR", Proc. of the SPIE: Algorithms for Synthetic Aperture Radar Imagery VII, Aerosense, April 2000, Orlando, Fla.
A. J. Kim, and "Nonparametric Estimation of Aspect Dependence for ATR", Proc. of the SPIE: Algorithms for Synthetic Aperture Radar Imagery VI, Aerosense, April 1999, Orlando, Fla.