Research On Array-based 3D Imaging And Enhaned Radar Imaging Techniques

Radar imaging is an important technology to obtain the targets information in both civil and military fields,and it is playing an irreplaceable role in developing the econo-my and strengthening defense.By reviewing the development course of radar imaging techniques,the thesis points out that the promotion of the image dimensionality and the improvement of the image quality are two essential trends of the future development of radar imaging.Accordingly,researches are carried out around two themes,i.e.ar-ray-based 3D imaging and enhanced radar imaging,aiming at producing more informative images with higher quality more efficiently.Chapter one and two discuss the research significance,review the research history and current status,and summarize the basic knowledge and some common issues of the following research.Chapters three to six present several specific problems around the two themes in detail respectively.Chapter seven concludes the whole thesis and shows some future directions.For array-based 3D imaging,the research are conducted in two aspects,i.e.3D re-construction and 3D imaging.For 3D reconstruction,the 3D reconstruction?also known as 3D modeling?here refers to the process of obtaining point cloud and 3D surface models of the target which is a classical problem in computer vision.In this paper,new solutions built upon array-based millimeter-wave holography are proposed.These methods also extend the conventional connotation of radar imagery.Series of recon-struction methods and techniques with the frequency interferometry method as their core are proposed for the“SISO-planar scanning”and the“SISO-cylindrical scanning”regimes,e.g.the methods of SNR promotion,phase unwrapping,sequential estimation,sub-aperture division and the technique of cylindrical-to-planar wave transformation.Both simulations and experimental results verify that the proposed methods are able to achieve high precision reconstructions of different targets with various shapes.It also shows that the proposed methods are of high precision,high efficiency,the ability to acquire the absolute depth information and the moderate penetration of clothes.For 3D imaging,aiming at fixing the“pain point”of lacking efficient imaging algorithms for the“scanning-MIMO array”regimes,several accurate and fast imaging algorithms are proposed for the“MIMO-planar scanning”and the“MIMO-cylindrical sanning”re-gimes based on the methods of spherical wave decomposition,nonuniform Fourier transforms and circular convolutions.These proposed algorithms extend the classical frequency domain imaging algorithms which are only valid under the“spherical wave-SISO”assumptions to the“spherical wave-MIMO”cases.According to the theo-retical analyses,compared with the recently widely used back projection-based imaging algorithms for“scanning-MIMO array”regimes,the proposed algorithms reduce the computational complexity from O?7?N ⁶?8?to O?7?N⁴log₂ N?8?.Simulations and experi-ments further verify the superiority of the proposed algorithms in imaging efficiency.For enhanced radar imaging,the research are also conducted in two aspects,i.e.sparse recovery-based and deep network-based enhanced radar imaging.For sparse re-covery-based enhanced imaging,since the linear imaging algorithms are of the intrinsic limitations of limited resolutions and high level side-lobes,kinds of off-the-shelf sparse recovery methods are utilized to achieve enhanced radar imaging for different ar-ray-based imaging regimes.Three specific sparse recovery algorithms belong to the“on-grid”type,“off-grid”type and the“gridless”type respectively are introduced.The“on-grid”sparse method is applied to a novel array-based standoff screening system.Compared with the non-enhanced linear imaging results,the enhanced images achieve better resolution and side-lobe performances.The“off-grid”type and the“gridless”type sparse recovery methods are adopted to tackle the previously proposed 3D recon-struction problems.These algorithms conquer the drawback that the interferometry-based method is only valid for single component signals and can achieve high precision 3D reconstruction for multi-layer targets.For deep network-based en-hanced imaging,since recent sparse recovery methods are usually time-consuming and unstable,it is the first time that the a deep network is used as an universal regressor to achieve enhanced radar imaging.It provides an entirely new solution to break through the current technical bottlenecks of enhanced radar imaging.The basic processing framework,the elementry solutions to complex-value issues and the training data gen-eration issues are firstly presented.Then the proposed method is tested under a typical enhanced radar imaging scenario.The results show that the proposed method is of ob-vious superiorities on resolution and side-lobe performance compared with traditional linear methods.It is also 2 to 3 orders faster than typical sparse recovery methods which makes the real-time enhanced imaging possible.Overall,theoretically,the thesis proposes several new solutions,new models and new methods for array-based 3D imaging and enhanced radar imaging.Practically,the works in the thesis can be directly applied to nondestructive examination,security in-spection,ISAR imaging,3D SAR imaging and so on.