Astronomical Image Reconstruction And Information Retrieval On Interferometric Synthetic Aperture With Sparse Baselines

Synthetic Aperture Interference Radiometer(SAIR)is a typical imaging system sampling signal in spatial frequency domain.It obtains the visibility function by complex correlation of pairs of received microwave signals from an antenna array and obtains the brightness temperature domain image through inverse Fourier transform of the visibility function.With the development of interference technology,the base interference array is developed toward a space-based interference array.In the field of astronomical interference,space-based interference arrays can effectively avoid influence of the Earth's ionosphere,especially for very low frequency observations;in the field of Earth observation,due to advantage of the long space array baseline,application fields are also aiming at high-precision targets and Trace and other directions from large-scale meteorological and oceanographic observations.Due to the limitation of the number of satellites,as well as system complexity,cost and other constraints,often adopted Sparse array.In the field of astronomical observation,imaging of all-sky and slowly-changing strong-source targets can be obtained from sparse arrays and time-sharing sampling by taking the sampling of the visibility function in the spatial frequency domain and then performing the brightness temperature inversion to reconstruct the brightness temperature of the observation area.But it does not work for time-varying targets.In the field of Earth observation,due to the complex ground environment,targets can be hardly detected by time-sharing measurements.In the case of sparse baseline,synthetic aperture image reconstruction and information extraction face two problems.On the one hand,due to insufficient sampling in the spatial frequency domain,traditional methods such as Fourier transform and G matrix are difficult to reverse the original image information.On the other hand,in the time-sharing mode,The affects on the imaging result by the variation of the brightness temperature of the time-varying source during the imaging period cannot be ignored,and traditional methods have difficulty detecting time-varying sources and reconstructing changing images.SAIR imaging technique acquires original brightness temperature(TB)image by Fourier transformation of measurements in the spatial frequency domain.According to the sampling principle,interval of uniform smapling derteminates the minimum aliasingfree field of view,while the biggest distance determinates the ressolution of the inversed image.When the sampling is sparse,if the minimum afFOV corresponding to the interval of sampling in spatial frequency domain is less than the scene,the inversed image is aliased and it is difficult to detect moving targets in an sequence of such aliased images.On the other side,when the sampling is not unform,due to the non-Dirac point spread function,moving targets are spreading all over the scene in the sequence of inversed images.When the actual baseline can not cover all the smapling points,there is aliasing.In the case of undersampling,the measurement equation is an undetermined equation,and the aliasing and spreading problems need to be solved in targets detection.To sovle the above problems of sparse array image inversion,two image inversion methods are proposesed: one is compressed sensing based sparse baseline synthetic aperture image reconstruction algorithm,and the other is Generative Adverarial Network based sparse sparse baseline synthetic aperture image reconstruction algorithm.To solve the problem of time-varying source information extraction from sparse arrays,a fixed-baseline sparse sampling motion point target detection algorithm is proposed,and a variable-baseline sparse sampling time-varying source image reconstruction algorithm.Details are as follows:1.To solve the problem of sparse sampling image reconstruction,a compressed sensing based sparse baseline synthetic aperture image reconstruction algorithm is proposed.The image is reconstructed by the compressed sensing algorithm,and the residual visibility function is further reconstructed by deconvolution method.Simulation experiments prove it can solve image aliasing problem under sparse sampling,obtain highfidelity brightness temperature image.2.To solve the sparse sampling image reconstruction problem of complex scenes,Generative Adverarial Network(GAN)based sparse sparse baseline synthetic aperture image reconstruction algorithm is proposed.By learning the characteristics of complex scene images and fitting the conditional distribution of complex scene image pixels,the mapping relationship between the inverse Fourier transform image and the reconstructed image is obtained.Simulation experiments verify that it can quickly and effectively reconstruct the complex scene brightness temperature image.3.To solve the problem of interference to target detection by artifact and aliasing in inversion images under fixed baseline sparse sampling,a motion point target detection algorithm based on mean difference of visibility function time series and deconvolution is proposed.The mean difference is used to eliminate the background component in the visibility function which makes the deconvolution method converge quickly.Then targets are reconstructed and detected by the CLEAN method.It is verified by simulation experiments that it has low false alarm rate and false alarm rate.4.Solve the problem of target blur and dynamic information loss in time-varying source reconstructed images under variable baseline sparse sampling,a sparse baseline synthetic aperture dynamic scene image reconstruction algorithm based on sparse constrain of direct product space of background and target variation is proposed.Simulation experiments shows it can accurately reconstruct image and get dynamic information.