Research On Feature-Enhanced Terahertz Radar Imaging

Because of the high resolution of terahertz imaging and the advantages of anti-stealth and anti-jamming,terahertz imaging has considerable potential in the field of radar imaging.This dissertation focuses on improving the imaging quality of terahertz radar and emphasizing the features of the target.The main work is as follows:In the first chapter,the background and significance of the topic are introduced,the research status of terahertz radar imaging and radar image feature enhancement at home and abroad is introduced,and the structure of the dissertation is given.The second chapter studies the basics of terahertz radar turntable imaging.Firstly,the observation model of turntable imaging is established,the projection-slice theorem is introduced,and the convolution back-projection algorithm is briefly deduced.From the perspective of spatial frequency domain,the Rayleigh resolution and the maximum unambiguous distance are deduced.Then the regularization-based turntable imaging method is introduced,and the regularization framework of turntable imaging is given from the perspective of solving inverse problems,and the statistical interpretation of regularization method is given.Because of the large amount of data captured by terahertz radar,in order to reduce memory overhead,the existing imaging model is improved,and a sparse matrix imaging model is proposed,which provides convenience for data processing.In this dissertation,the features of the turntable imaging target of terahertz radar are analyzed,and it is concluded that the target in terahertz band has features of both point-based type and region-based type.In the third chapter,a terahertz radar point-based feature-enhanced imaging model is established under regularization framework.According to the characteristics of strong scattering points of terahertz radar targets,a priori model of point sparseness is derived based on geometrical theory of diffraction.Then the relation between the value k of the regular term and the prior information is analyzed.The conclusion is that the smaller the k is,the better the scattering point can be preserved by the point-enhanced regular term.Then the optimization algorithm based on quasi-Newton method is deduced,and a series of simulated and real experiments are carried out.The simulation shows the frequency broadening,fast convergence,insensitivity to initial conditions and good anti-noise performance of the algorithm.The results of radar turntable imaging tests turned out to be satisfying.In the fourth chapter,a terahertz radar region-based feature-enhanced imaging model is established under regularization framework.Similar to the third chapter,the relationship between the norm and the prior is analyzed.It is concluded that the smaller the k is,the smoother the surface of the target will be and the edge of the region will be preserved.The problem that the speckle level cannot be suppressed by surface feature enhancement is pointed out and the point-surface feature enhancement is carried out.Based on this,a region weighted terahertz radar feature enhancement imaging model is proposed,and the results are verified by simulation.The convergence of the algorithm,sensitivity to initial conditions and noise immunity are tested by simulation.The results show that the algorithm has a good anti-noise performance,but the change of initial conditions may lead to a bad reconstructed image.The fifth chapter summarizes the work and points out the future research directions.