Research On Linear Array Synthetic Aperture Radar Three-Dimensional Sparse Imaging Technology

Linear array synthetic aperture radar(LASAR)has the three-dimensional(3D)imaging ability with high resolution,and has huge application prospects in several fields.Traditional algorithms mostly achieve 3D imaging based on the matched filter(MF)theory,they exist huge echo signal and insufficient imaging resolution.Therefore,this dissertation studies 3D sparse imaging via compressed sensing(CS)algorithms.Compared with traditional algorithms,CS algorithms obtain higher imaging resolution using partial echo signals.However,CS algorithms still use all scattering units in the imaging scene to construct measurement matrix and estimate scattering coefficients,leading to high dimensional measurement matrix operations and low computational efficiency.Besides,most CS algorithms consider all targets to match the scattering units accurately,they achieve sparse imaging under the On-Grid strategy.However,there inevitably exists positionerrors among targets and scattering units in the actual imaging scene,which results in decreased imaging quality.In addition,when setting the uniform linear array(ULA)in the LASAR system,the spacing between adjacent elements needs to satisfy the Nyquist sampling theorem,it must be smaller than half a wavelength to avoid grating lobes,making the huge elements and mutual coupling in the ULA,which also leads to the difficult engineering implementation.Thus,this dissertation conducts research in the following areas,its main work and innovations are:1.Aiming at the low computational efficiency due to high-dimensional measurement matrix operations of the whole imaging scene,the fast compressed sensing algorithm via adaptive threshold(FCSAT)is proposed.Based on the sparsity of echo signals in the range direction,echo signals are classified into several subclasses by the Fuzzy C-Means(FCM)algorithm to generate the extraction threshold,the imaging scene with higher echo signals' amplitude than the threshold is extracted as the target areas and used to conduct 3D sparse imaging.Then,to avoid the preset subclasses in the FCM algorithm,the target areas are extracted under the maximum interclass variance criterion by the OTSU algorithm.Both simulation and experimental results indicate that the FCSAT algorithm obtains higher imaging quality and computational efficiency than conducting 3D sparse imaging on the whole imaging scene,its computational efficiency can be improved higher than 10 times.Secondly,the fast sparse recovery algorithm via resolution approximation(FSRARA)is proposed.The FSRARA conducts sparse imaging to obtain every equidistant planar low-resolution imaging results quickly,and conducts image segmentation to extract their target areas coarsely.The target areas' low-resolution imaging results are performed the linear interpolation to obtain their high-resolution image,and conducted the secondary image segmentation to obtain more accurate target areas,the measurement matrix and scattering coefficients of target areas are built by their position information.Based on the prior distribution of target areas' scattering coefficients,sparse imaging model,and Bayesian information criterion,the cost function of scattering coefficients is established,and the minimum cost function is calculated to achieve the optimal estimation of scattering coefficients.Both simulation and experimental results show that the FSRARA compensates for the FCSAT algorithm's shortcomings in just using the sparsity along range direction,it achieves 3D sparse imaging with high efficiency and accuracy successfully.Its imaging quality and computational efficiency are higher than Orthogonal Matching Pursuit(OMP),Fast Iterative Shrinkage Thresholding Algorithm(FISTA),and Sparsity Bayesian Recovery via Iterative Minimum(SBRIM)algorithms.Especially,compared with the SBRIM algorithm with the highest imaging quality,its computational efficiency can be improved more than 250 times.2.Aimed at the decreased performance or failure estimation of the FSRARA due to the false targets,the fast Bayesian compressed sensing algorithm via relevance vector machine(FBCS-RVM)is proposed.The FBCS-RVM algorithm gives the scattering units in the imaging scene independent hyperparameters,their marginal likelihood functions are obtained based on the prior distribution of scattering coefficients,Bayesian probability criterion,and matrix decomposition,it extracts target areas by iterative optimization estimation.Under every iteration,according to whether have the maximum marginal likelihood function,scattering units are classified into target and background scattering units.Their marginal likelihood functions' extreme values are calculated to update their hyperparameters,estimation errors,and target areas,iterations terminate when reaching preset accuracy.Then,according to the position information of target areas,their measurement matrix,scattering coefficients,and cost function are established.The minimum cost function is calculated by iterative optimization estimation to achieve the optimal estimation of scattering coefficients.However,under high sparsity of imaging scene or low signal-tonoise ratio,there probably exists false targets which lead to the singular matrix's appearance during iterative estimation,resulting in the scattering coefficients' failure estimation.Thus,the inverse operation on the singular matrix is replaced with the truncated singular value decomposition(TSVD)algorithm to achieve the high-accuracy estimation of scattering coefficients.Both simulation and experimental data indicate that the FBCS-RVM algorithm avoids the failure estimation due to false targets,it achieves 3D sparse imaging with high efficiency and accuracy.Its imaging quality and computational efficiency are higher than the OMP and SBRIM algorithms.Especially,compared with the SBRIM algorithm,its computational efficiency can be improved more than 200 times.3.Aimed at the decreasing imaging quality due to the position-errors,base on the Off-Grid strategy,this dissertation proposes the sparse recovery algorithm via adaptive grids(SRAG).The SRAG algorithm obtains the On-Grid imaging results and extracts target scattering units by image segmentation.Then,it is considered that target scattering units with non-zero scattering coefficients of adjacent scattering units have position errors,they are defined as unfocused scattering units.According to whether the distance among unfocused scattering units is smaller than scattering unit spacing,the search-areas of every unfocused scattering unit are set differently.The scattering unit with minimum scattering coefficient residual error in the search areas is extracted and used to update unfocused scattering units' positions,optimize the measurement matrix,and update their scattering coefficients.After gradually updating unfocused scattering units and narrowing their search areas,iterations terminate when the number of unfocused scattering units is fixed.Both simulation and experimental results prove that the SRAG algorithm decreases the position-errors effectively,it improves the estimation accuracy of targets' positions,scattering coefficients and imaging quality than both OMP and SBRIM algorithms.4.Aimed at the huge elements and mutual coupling in the ULA,the joint sparse recovery(JSR)algorithm is proposed for Coprime Adjacent Array(CAA)SAR 3D sparse imaging.Firstly,the 3D sparse imaging model of CAA-SAR is built,which decreases the array elements and suppresses the mutual coupling successfully,it also leads to false targets in the imaging results.Aimed at the above problem,the joint sparse recovery(JSR)algorithm is proposed.The proposed algorithm decomposes the CAA into two uniform sparse subarrays with different spacings,it conducts sparse imaging to obtain the imaging results of CAA and its subarrays.The target areas' imaging results of CAA and its subarrays are extracted by image segmentation,the above three target areas' imaging results are performed image fusion by the discrete wavelet transform to eliminate the false targets and obtain more accurate imaging results.Both simulation and experimental results indicate that the JSR algorithm obtains higher imaging quality and computational efficiency with fewer elements than ULA-SAR 3D sparse imaging,its elements can be less than 12.5%of ULA.Compared with CAA-SAR 3D sparse imaging,the JSR algorithm eliminates the false targets effectively,and obtains higher imaging quality and computational efficiency.