A Singular Value Decomposition Based Approach for Classifying Concealed Objects in Short Range Polarimetric Radar Imaging

Abstract

In current research one of the main challenges in short range synthetic aperture radar (SAR) is electrically small structures and objects, which tend to unclear reinforced or through the wall objects, object orientation angle, and obscure contribution to extract the position of concealed multiple small objects. In this paper, ultra-wide-band (UWB) polarimetric radar was used to study reinforced objects and for estimation of object angle at short range. Electrically small 1D periodic mesh, 2D periodic meshes and differently oriented small objects or meshes could not be distinguished in conventional SAR images. A radar system with transmit and receive antennae mounted on a two dimensional scanning grid was used. The aim is non-destructive testing of built structures, in concrete slab manufacturing and for use in the renovation process. UWB short range radar data and images corresponding to different polarization states were analysed by using singular value decomposition (SVD). To perform decomposition, the proposed approach applies SVD to image data matrices produced from the back projection algorithm (BPA) to classify the different objects and identify the object angle. Then, sets of singular-components of different polarization states are analysed to classify objects. Also, the BPA algorithm is performed to construct the object images from the polarimetric radar signals. The object reflection varied with the polarimetric state of the UWB radar, which contributes to different object signatures (i.e., object intensity) since the object signature depends on the orientation, the size, and the number of objects. Object orientation with respect to the radar system and object anisotropy could be determined from the ratio of the different polarimetric singular-components. This proposed complex data analysis method demonstrates the usefulness of the SVD using BPA in extracting more information about and for classifying an object.