Research On Near-Field Three-Dimensional Imaging Method Based On Compressed Sensing

Near-field three-dimensional(3D)imaging can provide 3D position and scattering characteristic information of scattering points.Because of its high imaging accuracy and ability to overcome the shadow effect of two-dimensional imaging,it has important application significance in the fields of target scatter imaging diagnosis,security check,and through-the-wall detection.Traditional imaging methods based on matched filtering will face the following problems when improving resolution: high sampling rate,difficult processing of massive data,and broadening of main lobe.Compressed sensing(CS)can realize high-quality imaging on sparse sampled data by utilizing the sparse prior information of the target,so it shows great potential in reducing sampling rate of the imaging system,reducing data volume and improving imaging quality.Based on near-field 3D imaging theory and CS theory,this dissertation focuses on the near-field 3D imaging methods based on CS for purpose of improving imaging quality.According to the structure of near-field 3D imaging system,the dissertation analyzes the mechanism of near-field 3D imaging in four modes: interferometric imaging,planar scanning imaging,circular scanning imaging and through-the-wall imaging.The main purpose is to solve the problems of CS imaging in the above four modes.Main contributions and innovations of the dissertation are as follows:1.Multi-channel joint sparse reconstruction interferometric 3D imaging method.Traditional independent CS based imaging method cannot guarantee the number and position consistency of scattering points in each channel,which leads to pixel mismatch and non-uniform focusing in height information extraction.An interferometric 3D imaging method based on multi-channel joint sparse reconstruction is proposed.This method constructs a multi-channel joint sparse imaging model based on prior information with the same sparse support between channels,and takes the mixed norm of multi-channel scattering coefficients as the global sparse consistency constraint.In this method,atoms are selected by calculating the mixed norm of the residual of multi-channel and the inner product value of the corresponding observation matrix,and the scattering coefficients of each channel are solved within the same support set to achieve the consistency of the number and position of scattering centers in each channel.Simulation and experimental results show that,compared with independent channel imaging,this method can achieve high quality imaging with a small number of samples.2.Near-field 3D imaging method of planar scanning based on CS.Aiming at the problem of defocusing imaging caused by phase drift of echo due to long time measurement,a method of compressed sensing 3D imaging based on phase error estimation and compensation is proposed.By analyzing the relationship between the echo data with phase error and the observation model,a joint imaging and phase error correction model based on non-quadratic regularization was constructed.During the imaging process,the error term was estimated and the dictionary was corrected to make the dictionary atom accurately match the echo signal and realize high focus imaging.Simulation and experimental results show that the method can effectively deal with various phase errors and has better imaging focusing ability than traditional autofocus imaging methods.To solve the problem of complex computation and large memory usage of explicit dictionary CS imaging,a near-field 3D imaging method based on Range Migration Algorithm(RMA)imaging operator and Fast Gaussian Grid Non-uniform Fast Fourier Transform(FGG-NUFFT)is proposed.Starting from the original data domain,this method uses the forward and inverse transforms of RMA to construct the dictionary,which improves the computational efficiency and reduces the memory demand.The FGG-NUFFT is used to improve the interpolation accuracy and further accelerate the imaging speed.Simulation and experimental results show that,compared with the traditional CS imaging method,this method can not only guarantee the imaging quality,but also greatly improve the imaging efficiency and reduce the memory requirements.3.Near-field 3D imaging method of circular scanning based on CS.In order to solve the problem of high sidelobe and more spurious points caused by the anisotropy of target scattering in traditional circular imaging,a layered 3D imaging method based on sub-aperture joint sparse constraint is proposed.Firstly,the sub-aperture imaging method is used to reduce the influence of target scattering anisotropy.Secondly,the redundancy and correlation between sub-aperture echoes are fully considered,and a sub-aperture joint sparse imaging model based on mixed norm is constructed.It ensures the uniqueness of the position of the scattering points,realizes the sidelobe and spurious point suppression and compensates for the defect that the subspace causes the resolution to decrease.Simulation experiments show that this method can effectively improve the quality of circular imaging.Aiming at the problem of large volume of data and low imaging efficiency for cylindrical scanning imaging,a 3D imaging method based on RMA and Back Projection(BP)imaging operator is proposed.In this method,3D imaging is decomposed into two 2D imaging along the altitude direction and the horizontal direction.Rapid decoupling and focusing of the altitude direction are realized through RMA imaging operator in frequency domain,precise focusing of the horizontal plane is realized through BP imaging operator in time domain,and the number of altitude antenna array elements is reduced through CS sparse reconstruction.Simulation experiments show that the method can achieve high-efficiency and high-quality imaging with a small amount of observation data.4.3D imaging method of through-wall based on strong clutter suppression.Aiming at the problem that wall clutter seriously affects the imaging quality,two 3D imaging methods based on strong clutter suppression are proposed.The first method is a through-the-wall 3D imaging method of based on subspace decomposition and CS.This method defines a new wall subspace division criterion by the left singular matrix variance stable level,avoiding the use of continuous subspaces to characterize the wall clutter,so that the division of the wall subspace in the echo signal is more accurate,and the residual wall component in the target subspace is effectively removed.The basis pursuit de-noising algorithm is used to achieve sparse three-dimensional reconstruction while suppressing noise.Simulation experiments show that this method can effectively improve the quality of through-the-wall imaging.The second method is a 3D through-the-wall imaging method based on low rank sparse constraints.The method takes advantage of the low rank of the wall signal and the sparseness of the target signal.By applying a norm constraint on the matching of different signals,a joint regularized imaging model including the wall signal and the target signal is constructed,clutter suppression and 3D imaging are transformed into the optimal solution of the mixed norm.In the process of imaging,the optimal separation of wall signal and target scattering is realized by iterative discriminant method,which achieves the purpose of effective clutter removal and high quality imaging.Simulation and experimental results show that,compared with the traditional through-the-wall imaging method,the method has strong clutter suppression ability and good imaging quality.