Time-Varying Wideband Interference Mitigation for SAR via Time-Frequency-Pulse Joint Decomposition Algorithm
Jia Su,Mengru Xi,Yanyun Gong,Mingliang Tao,Yifei Fan,Ling Wang
DOI: https://doi.org/10.1109/TGRS.2022.3213318
IF: 8.2
2022-01-01
IEEE Transactions on Geoscience and Remote Sensing
Abstract:Wideband interference (WBI) may severely affect the imaging quality of synthetic aperture radar (SAR) systems. Since it highly overlaps with useful signals in the 1-dimensional (1-D) time or frequency domain, the existing WBI mitigation methods usually transform 1-D echoes into a 2-D transform domain. However, they usually suffer from a model mismatch, which results in the loss of the useful signal. To tackle this problem, a novel algorithm combining time-frequency-pulse (TFP) joint characteristics and robust principal component analysis (RPCA) is proposed for WBI mitigation. The TFP joint feature of SAR echo is introduced for interference mitigation for the first time. We first transform the SAR echoes into the time-frequency domain and construct a new TFP matrix by reshaping the STFT matrices between adjacent pulses. In terms of WBI-occupied SAR echoes, the short-time Fourier transformation (STFT) in adjacent pulses can be modeled as a combination of a low-rank part (i.e., useful SAR echoes) and a sparse counterpart (i.e., WBIs), which well fits the assumption of RPCA. Then, the TFP matrix is decomposed into the useful signal TFP matrix and the WBI TFP part by taking full advantage of the low-rank and sparse properties. Finally, the WBIs can be reconstructed and subtracted from the echoes to realize interference mitigation. Experimental results on both simulated and measured datasets show that the proposed algorithm not only suppresses WBIs effectively but also preserves useful information as much as possible.