Novel 3D Interferometric Inverse Synthetic Aperture Radar Imaging Techniques for Non-Cooperative Target Recognition

In modern defense and surveillance, accurately identifying non-cooperative targets with unpredictable movements is a critical challenge. While two-dimensional ISAR can produce high-resolution images, it often lacks the three-dimensional detail necessary to fully characterize complex objects. 3D Interferometric Inverse Synthetic Aperture Radar (InISAR) imaging addresses this limitation by leveraging interferometric phase differences to reconstruct detailed 3D images. This doctoral thesis introduces novel methodologies to advance 3D InISAR imaging, with a focus on improving the recognition of non-cooperative targets through three key contributions. First, it introduces the Modified Polarimetric CLEAN algorithm, which improves the quality of scattering center extraction, in combination with fully polarimetric approaches—Coherence Optimization and Polarimetric Matched Projection—to refine phase estimation in 3D InISAR systems, leading to more precise and reliable reconstructions. These methods are validated through simulations of a point-target and a real-world dataset, demonstrating their effectiveness while also highlighting potential limitations. Second, the thesis addresses the challenge of phase unwrapping by proposing a multi-frequency technique that mitigates phase ambiguity, improving reconstruction accuracy. This approach is particularly effective for man-made targets, where traditional unwrapping methods often fail. Finally, it contributes to Automatic Target Recognition (ATR) by exploring artificial intelligence methods, with a focus on Transformer architectures. These models capture complex relationships in multidimensional data, increasing robustness and improving target recognition accuracy. By integrating AI with advanced radar signal processing, this approach enhances system adaptability and reliability.