Tensor-Integrated Acceleration Framework for Sparse-Based WAS-GMTI in Multichannel Airborne Radar Systems

Wide-Area Surveillance Ground Moving Target Indication (WAS-GMTI) persists as a critical research direction in the airborne radar field due to its great potential in ultra-long-range situational awareness missions. Nevertheless, the pivotal clutter-suppression component employing Space-Time Adaptive Processing (STAP) has long been plagued by performance degradation due to training sample scarcity in heterogeneous terrain environments. In this article, aimed at addressing this challenge and further extending the WAS-GMTI technique by integrating state-of-the-art Sparse Recovery (SR) theory, we propose a general Sparse-based WAS-GMTI (S-WAS-GMTI) framework for multichannel airborne radar applications. The framework was originally developed using a tensor representation to ensure both computational efficiency and scalability to high-dimensional echo data processing. Next, to face real-time deployment requirements, two Reduced-Dimensional (RD) versions S-WAS-GMTI framework are systematically designed to optimize computational efficiency. Simultaneously, given the high computational cost associated with SR optimization, we also propose Fourier-based overcomplete dictionary construction methods and a fast Fourier Transform-based tensor Sparse Bayesian Learning (SBL) method, resulting in several orders of magnitude improvement in solving efficiency compared to traditional SBL SR methods. More significantly, diverging from our prior work, we employ our recently developed advanced multi-channel airborne radar system to experimentally validate the proposed S-WAS-GMTI. Extensive real-measured datasets conclusively demonstrate the superior performance of S-WAS-GMTI relative to the existing WAS-GMTI processing.