Adaptive Object Classification Using Complex SAR Signatures

Firooz Sadjadi; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2016, pp. 33-37

This paper addresses the problem associated with the classification of signatures of objects obtained by coherent sensors whereby the signatures are complex valued. Individual phase and amplitude component of a signature are combined optimally and the resulting fused signature is used in a sparsity-based learning classifier. The results of application of this approach are then compared with the corresponding results using only the amplitudes of the signatures. To test the concept public-domain radar signatures of several land vehicles obtained at different aspect angles are used. The performance improvement, based on confusion matrices, is shown to be significant when both phase and amplitudes are used.