Accelerated L 1 =svd Deconvolution Approach for Real Aperture Radar Super-resolution Imaging

Xingyu Tuo,Yin Zhang,Yongchao Zhang,Yulin Huang,Jianyu Yang

DOI: https://doi.org/10.1109/radarconf2248738.2022.9764344

2022-01-01

Abstract:Recently, the l(1) - svd deconvolution approach has been proposed for real aperture radar (RAR) super-resolution imaging in low signal-to-noise ratio (SNR) condition. However its real-time ability is restricted due to complex matrix-matrix operations. We further develop this approach by accelerating it in this paper. Firstly, we uses the truncated unitary matrix to replace the truncated convolution kernel matrix to compress the convolution model. The reduction of matrix dimensions brings acceleration of matrix-matrix multiplications. Aiming at the inverse operations in this deconvolution method, ShermanMorrison formula is exploited to further acceleration. After the above processing, the total computational complexity is reduced from N-3 to kN(2). Simulation and experimental data prove that the proposed approach offers significant computational savings without loss of imaging performance.