Research On SAR Imaging Technology Based On Sparse Reconstruction

Since high resolution SAR imaging technology is widely used, radar needs to transmit a wide-band signal. In the Nyquist sampling theorem framework, it results in high sampling rate, huge data volume and the difficulties of real-time processing and hardware implementation. However, the echo signal of radar itself has sparse property, which makes it possible to break the restriction of Nyquist sampling theorem and use a sparse signal to recover the SAR image. The theory of Compressive Sensing(CS) and Matrix Completion(MC) are among the common methods that can deal with sparse signals. They have obvious advantages in SAR imaging. At the same time, the current design methods of observation matrix and reconstruction algorithms in the CS remain to be improved, so optimizing them with the gradient descent method has a certain research value.This thesis focuses on the methods of sparse SAR imaging. The main work can be summarized as follows:1. The model of SAR echo signal and Range-Doppler method are introduced. Then, the validity of this method is proved by point targets model and real data.2. This thesis have a research on the model of SAR imaging based on CS and analyze the sparsity of radar echo signal. On the basis of this, how to formulate the sparse basis matrix and the observation matrix is discussed. Then, the problem of SAR imaging based on CS is solved by translating it into mathematical optimization model. Point targets simulation of this method is conducted in range dimension, azimuth dimension and two dimension. The results show that this method can both reduce the amount of sampling data and improve the resolution by a large margin. The results of real data also prove its validity.3. On the basis of CS SAR imaging, the observation matrix and the reconstruction algorithm are optimized. At first, the gradient descent method is introduced briefly. Then, the factor showing the performance of observation matrix is analyzed. And weuse the gradient descent method to optimize this factor so that the observation matrix is also optimized. This new method shows higher reconstruction quality compared to previous method in the simulation. At last, the algorithm of OMP and SL0 are introduced. Based on the OMP and the gradient descent method, a new algorithm is discussed in this thesis. It has a faster convergence rate. In the simulation, the new algorithm is better than the original OMP in terms of reconstruction time.4. In the research on SAR imaging method based on MC, we have a brief explanation on the low-rank property of radar echo signals. Then, the model of SAR imaging and the method to formulate the down-sampling set are introduced, which are applied in SAR imaging with undersampled data in range cells. In the case of noise, we conduct both point targets and real data simulations, the results show that this method can both reduce the amount of sampling data and reflect the scattering information of point targets exactly. It has the capability of anti-noise, too.