Research On Multidimensional Radar Imaging Methodology By Means Of Compressive Sensing

Imaging radars can retrieve the scattering reflectivity map of an interested target from measurements of scattered electric fields, and they have played an important role in application, such as target recognition. In order to provide a more comprehensive and refined descriptions of scattering characteristics of interested targets, multidimensional radar imaging including passive bistatic radar imaging, high-resolution fully polarimetric radar imaging and three-dimensional radar imaging has emerged as the times required. The newly-built Compressive Sensing(CS) theory gives a brand new framework for signal acquisition, which allows a sparse signal to be recovered exactly from measurements sampled at a rate far below the Nyquist sampling one. As a consequence, if we apply the CS to radar imagine field, it is promising to solve problems faced by traditional imaging radar, such as high sampling rate, large amount of data, and radar imaging by using incomplete data. Aiming at improving the capability of imaging radar power and by taking radar imaging via CS as the mainline, this dissertation is dedicated to the techniques needed in the passive bistatic Inverse Synthetic Aperture Radar(ISAR) imaging, high-resolution fully polarimetric ISAR imaging and three-dimensional radar imaging.Chapter 1 firstly introduces the research background and significance of this dissertation; and then it surveys the state-of-the-art development on the CS theory and CS-based radar imaging methods; subsequently, it points out some problems required to be tackled in three kinds of imaging radar; finally, the main research work and the organization of this dissertation is outlined.Chapter 2 describes the basic principle of ISAR imaging based on CS. Firstly, it briefly recalls the mathematical model of CS theory; then by using stepped-frequency waveform, it derives the signal model of ISAR imaging and applies the CS to the ISAR imaging; two sparse measurement schemes are discussed subsequently, as well as the selection of reconstruction algorithm and dictionary refinement parameter, and in the following the high-resolution ISAR image can be obtained by solving an optimization problem; finally, experimental results by simulation data and real data are shown to validate the effectiveness of this imaging method. This chapter serves as groundwork for the following researches.Chapter 3 studies the passive bistatic ISAR imaging via CS, which is conducted to reduce the grating lobes and improve the resolution of the conventional passive ISAR images. Firstly, the signal model for passive bistatic ISAR exploiting the illuminator of opportunity is derived, which finally can be reformulated as a matrix form suited for CS framework from the perspective of statistical average; by solving an optimization problem under the constraint of signal sparsity, scattering centers in passive ISAR image can be recovered, and the completed data can then be calculated based on the signal model and the recovered information of scattering centers; after this step, the ISAR image can be obtained by applying the traditional range-Doppler imaging algorithm to the estimated complete data, which is comparative with that obtained by range-Doppler imaging algorithm to the measured data; furthermore, by utilizing the recovered parameters of scattering center and the signal model again, we can easily extrapolate the received signal in a wider frequency/slow-time domain, which surely will improve the resolution of the ISAR image; finally, both simulation and experimental results are shown to demonstrate the validity of the proposed approach.Chapter 4 proposes a novel high-resolution fully polarimetric ISAR imaging method by means of CS. Firstly, motivated by the fact that the fully polarimetric ISAR images share the same sparsity support over the polarimetric channels, i.e., joint sparsity, ISAR imaging is cast as a two-dimensional multiple measurements sparse signal reconstruction problem; then we define two mixed norms to characterize this joint sparsity property, and use continuous Gaussian functions to approximate these two mixed norms; subsequently, fully polarimetric ISAR image are constructed by the recovery algorithms aiming at minimization of these two mixed norms; finally, results from simulation data and measured data from anechoic chamber and field show that the advantages of the proposed method: it can provide high-resolution ISAR images with limited measurements and the number together with the positions of the scattering centers are aligned in polarimetric channels, which is beneficial for the further target recognition application.Chapter 5 develops two three-dimensional radar imaging methods based on CS: interferometric ISAR imaging and three-dimensional turntable ISAR imaging. For the former one, we consider two ISAR images from the baseline corresponding to the two antennas share the same sparsity support, and thus we can define two kinds of global sparsity inspired by the idea in Chapter 4; then we can construct high-quality ISAR images by solving the optimization problems under the constraint of global sparsity definition, and three-dimensional reconstruction resulted can further be generated by interferometry processing between the two ISAR images. For the latter one, conventional vector-CS method usually converts the three-dimensional data into a long vector along with a large dictionary, and it imposes a huge memory storage and computational burden. By exploiting the inner structure of measured three-dimensional data, we propose two efficient CS based imaging methods, namely dimension-reduced CS and tensor CS. Dimension-reduced CS reformulates the three-dimensional data model into a matrix form first, then applies the two-dimensional CS imaging technique to obtain the ISAR map, and next rearrange the two-dimensional ISAR map into three-dimensional volume, i.e., three-dimensional imaging result; tensor CS deals with the three-dimensional data directly with the help of tensor algebra, and a new reconstruction algorithm named tensor-SL0 is also proposed. These two imaging approaches can speed up the imaging processing procedure with respect to the vectorized CS one, reduce memory usage and computational complexity, and are promising to be applied to radar imaging of large targets. Experimental results by using simulation data and data generated by the EM code are finally shown to demonstrate that the proposed techniques can form three-dimensional images of the target efficiently.Chapter 6 concludes the whole dissertation, and presents the outlook of the future work.