Research On Array Radar Forward-Looking Imaging Technique Based On Wavefront-Modulating

Radar imaging technology is an information acquisition method that is less restricted by natural conditions and can work in all-day and all-weather conditions.Radar imaging technology has been widely applied in military reconnaissance and civil fields and how to illustrate higher imaging resolution in forward-looking scene is always ambitions of researchers.Aiming at array radar forward-looking imaging,this thesis focuses on investigating radar forward-looking imaging technique based on wavefront-modulating,expecting to provide theoretical support for the practical application of array radar forward-looking and superresolution-imaging techniques in practical imaging scenes.A series of novel theories,methods and imaging models are proposed in this thesis.Firstly,the backgrounds and significances of this thesis are introduced,the existing array radar forward-looking imaging techniques are sorted out,and the advantages and disadvantages of different imaging techniques are summarized.Moreover,the developing status of radar wavefront-modulating forward-looking imaging technique are introduced,and some fundamental and urgent problems are summarized.The resolution is a basic index to evaluate the radar imaging performance.Aiming at the lack of method for resolution measurement in radar wavefront-modulating forward-looking imaging technique,the algorithm-related point spread function(PSF)is defined to quantitively measure the resolving ability and regularity of radar wavefront-modulating forward-looking imaging technique.And the wavefront entropy is also defined to quantitively measure the fluctuation of radiation field.The essence of resolution improvement is revealed,that is for different radar imaging systems,the resolution can be improved by increasing the wavefront entropy.Further,the resolution measurements of traditional array radar imaging technique and wavefront-modulating forward-looking imaging technique can be incorporated into a unified framework by wavefront entropy.Utilizing the parametric imaging model can bring better imaging quality for radar wavefront-modulating forward-looking imaging technique,however,the discretization in modeling will lead to dictionary basis mismatch in practical imaging scenes.Based on the first-order Taylor expansion of the reference matrix and the statistical property of the elements of perturbation matrix,the directional-fission dynamic grid and all-around-fission reweighted dynamic grid are proposed,respectively.And the theoretical expected Cramér-Rao bound(ECRB)of proposed method is derived.Moreover,the azimuthal superresolution compared to real aperture imaging is obtained in the outdoor experiment.The discretizing modeling of radar wavefront-modulating forward-looking imaging technique will lead to grid mismatch,which can be well avoided by resolving the target in the continues field.In this thesis,a wavefront-modulating forward-looking imaging technique based on the continuous compressed sensing(CCS)method is proposed,which uses the one-dimensional atomic norm minimization(ANM)to solve the positions and amplitudes of scatterers in continuous field and improves the imaging quality.And the azimuthal superresolution compared to real aperture imaging is obtained in the simulating experiment.Obtaining sparse array in radar imaging could largely reduce the complexity of system,however,it will lead to wider main lobe and higher side lobes.A computation-aided sparse array radar imaging technique is proposed based on the matrix completion and CCS and imaging results with thinner main lobe,higher side lobes and continuous solutions are obtained,so that the imaging quality is greatly improved.Another problem brought by the use of parametric imaging models is the complexity of the solution.To solve this problem,a super-resolution imaging method based on regression-type deep convolutional networks is proposed in this paper.The Bernoulli-Gaussian sparse distribution mask is used to generate radar scattering.The characteristic imaging target dataset solves the problem of the current shortage of radar imaging datasets,and the dataset generation method can also be extended to other radar imaging systems.The trained imaging network can improve the range and azimuth resolutions,and the imaging time is greatly reduced,which can achieve high-quality imaging of the target in near real time.Nonlinear mapping transformations between point spread functions are learnt to achieve super-resolution imaging.The signal-to-noise ratio(SNR)of the radar echo has a crucial influence on the imaging quality.The signal of traditional radar wavefront-modulating forward-looking imaging is incoherent,and the imaging signal lacks power gain in the transmitting,receiving and processing procedure.In this thesis,a coherent-detecting and incoherent-modulating forward-looking imaging model is established,in which the pulse compression is used to detect the target coherently,and the post-random-modulation method is used to modulate the wavefront incoherently.Finally,the superresolution imaging result of the actual complex target is realized with low SNR.Unlike the classical imaging model which assumes the target as a set of discrete points,in this thesis,a novel radar wavefront-modulating forward-looking imaging model based on the attributed scattering model(ASM)is established.The ASM-based MCI equation contains three kinds of reference matrices corresponding to the point-scatterers(PSs),the line-segment-scatterers(LSSs)and the rectangular-plate-scatterers(RPSs),respectively.Hence the ASM-based imaging model could resolve richer information about the object geometries,which improves the imaging quality towards actual objects and promotes the application of wavefront-modulating forward-looking imaging technique.In the final,the main contributions of this thesis are summarized and aspects worth to be further investigated are enumerate.