Loading...

High-Dimensional Sparse Representation in ISAR Imaging

Nazari, Milad | 2021

487 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 53639 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Bastani, Mohammad Hassan
  7. Abstract:
  8. Sparse representation and compressed sensing have been widely used in various fields, one of the most popular of which is ISAR imaging. Inverse Synthetic Aperture Radar (ISAR) provides an electromagnetic image of the target, which is mainly used to identify and classify targets. In some applications, recognizing targets from a 2D image can be difficult and error-prone. One idea to deal with this problem is 3D ISAR imaging. The most widely used method of ISAR imaging is direct method based on Fourier transform. This method requires the measurement of radar data with high measurement density in 3 directions, which increases the data collection time and volume, which is the main problem of these methods. Our goal in this dissertation is to use sparse representation and compressed sensing to solve this problem.For this purpose, the problem of sparse representation in high dimensions has been formulated using the tensor theory. Also, the existing methods for solving it are discussed and a new method of High-Dimensional Smoothed l_0 (HDSL0) is presented, in which an approximation of l_0 norm with the sum of Gaussian functions is used for sparse recovery of high-dimensional data. To improve the performance of the HDSL0 algorithm, the MRe-HDSL0 algorithm is also proposed, which is a regularized and modified form of HDSL0 and uses a hard thresholding step inside its inner loop. In another part of the dissertation, the scattering point model of the target for 3-dimensional ISAR imaging based on compressed sensing is formulated and we have tried to reduce the imaging cost by using high-dimensional sparse reovery methods.According to the numerical simulations, using the proposed methods for high dimensional sparse recovery, the computational load and recovery time are reduced and the recovery quality is increased. In addition, the recovery quality for MRe-HDSL0 was 10, 9, and 8 dB higher in 1D, 2D, and 3D cases, respectively, compared to HDSL0, under the same test conditions. In addition, for radar simulations, we first used the SL0 algorithm and its modified version called MSL0 (which is also presented in this paper) for ISAR imaging based on the scattering point model of a simulated radar data. Based on the obtained results, the reconstruction of ISAR images has been done with appropriate quality. The proposed algorithms for high-dimensional sparse recovery algorithms have also been used to reconstruct 3D ISAR images of two artificial targets, based on the scattering point model. The results indicate that the quality of the images reconstructed using MRe-HDSL0 was better than other simulated methods
  9. Keywords:
  10. Compressed Sensing Radar ; Inverse Synthetic Aperture Radar ; Three Dimensional Radar Imaging ; High-Dimensional Sparse Representation ; Multi-Dimensional Signal Processing

 Digital Object List

 Bookmark

No TOC