A Distributed Cooperative SAR Three-dimensional Imaging Method
Dou Sun,Shiqi Xing,Bo Pang,Yongzhen Li,Xuesong Wang
DOI: https://doi.org/10.1109/piers-spring46901.2019.9017355
2019-01-01
Abstract:To obtain high-resolution and three-dimensional SAR imaging results, traditional three-dimensional SAR requires dense radar sampling in both azimuth and elevation, which requires a very large collection time and storage space, and may be very expensive in practice. In recent years, distributed collaboration technology has received extensive attention, especially the rapid development of UAV cluster technology. UAV cluster technology has the characteristics of low cost, large scale and high intelligence. Sparse sampled SAR data with multi-pass, large-baseline and wide-angle can be obtained by cooperative flight of multiple small UAVs, and the three-dimensional SAR imaging results with high resolution are expected. However, because of the strong maneuverability of UAV platform and the need to adjust flight trajectory to avoid obstacles, the Scanning trajectory of target through UAVs is not strictly regular, but a random curve. Therefore, k-space samples will be wide-ranging, non-uniform and sparse, resulting in characteristics of coupling and space-varying. Such data cannot be processed by traditional three-dimensional imaging methods. In this paper, a distributed cooperative SAR three-dimensional imaging method is studied. We first develop signal model for distributed cooperative SAR. Sub-aperture division is carried out based on the equal-interval division method. Then, the non-uniform sampling data in k-space is interpolated by Non-uniform FFT algorithms. Considering the sparsity of samples, the imaging problem is modeled as a sparse reconstruction problem. Finally, a fast algorithm of Lp-norm regularization is used for three-dimensional sparse imaging. The experimental results show that the proposed method not only can obtain high resolution three-dimensional imaging results for distributed cooperative SAR, but also has high efficiency and low computational complexity.