Array Configuration And Image Reconstruction Of Mirrored Interferometric Aperture Synthesis Radiometer

Interferometric aperture synthesis (IAS) is an interferometry, which uses a sparse array to replace a real aperture antenna and be considered as an alternative to the real aperture technique. It excludes the disadvantage of the ponderousness and heaviness of the real aperture antenna and can cover a large field of view without mechanical scanning, hence providing broad applications. With the increasing requirements of practical applications, it is revealed that IAS has a complicated system structure and complicated signal processing. Recently, a novel improved interferometry, mirrored interferometric aperture synthesis (MIAS), has been proposed, which is related to the two-element interferometry and the Lloyd's mirror interferometry. MIAS measures the cosine visibilities of the observed scene indirectly and the brightness temperature image can be reconstructed by signal processing. MIAS reduces system complexity and provides a better performance on spatial resolution compared with IAS. However, there are many problems should be solved in MIAS. For example, there is not a feasible sparse array for MIAS, which brings in the problems of baselines missing and the transmatrix's rank defect. And the image reconstruction of MIAS needs for a further study. In this thesis, the main aspects are studied as follows:(1) Array configuration of one-dimensional (1-D) MIAS is studied. According to the imaging principle of MIAS, there principles of array configuration optimal design are proposed:obtaining the longest max-baseline with no baseline missing, getting the maximum rank of the trans-matrix and having the fewest antennas. The second principle is proved to be the key of the optimization. Therefore, an optimization model for array configuration design of 1-D MIAS is established, and some optimal array configurations are presented through search algorithms. Simulations show that 1-D MIAS has a max-baseline almost twice that of 1-D IAS with the same array size, hence providing almost 100% higher spatial resolution. To achieve the same max-baseline, although the reduction in the number of antennas in 1-D MIAS is not much distinct, the number of analog-to-digital converters (ADCs) and correlators are reduced by 50% since the cross correlation is real and I/Q demodulation is not required, which reduces the system complexity.(2) Array configuration of two-dimensional (2-D) MIAS is studied. According to the imaging principle of 2-D MIAS, an optimization model for array configuration design of 2-D MIAS is established. Two aspects are proposed to reduce the complexity of the 2-D MIAS optimization problem. One is using a novel approach to design 2-D MIAS array. In the approach, many 2-D MIAS sparse arrays are found using a sieving method proposed by this thesis, then the characteristic of these arrays is analyzed. According to the characteristic, extra constraint conditions are attached to the optimization model which can reduce the complexity of the problem. The other is finding an efficient search algorithm. A modified group search optimizer (GSO) is proposed to be utilized in array optimization problems. Combining the above ideas, a method based on the modified GSO is proposed to sovle the 2-D MIAS optimization problem. By the method,2-D MIAS arrays with high sparsity can be obtained quickly. Moreover, a new type of array, double L shaped array, is presented, which has good practicality. Simulations show that a double-L shaped array in MIAS achieves higher spatial resolution and has lower system complexity compared with a U shaped array in IAS.(3) An image reconstruction method based on partial least-squares (PLS) regression is proposed. According to the imaging principle of MIAS, the image reconstruction model based on the impulse response matrix or the G operator is presented. The image reconstruction is considered as a linear regression problem and a modified PLS method is proposed, which is more appropriate than the original PLS method in solving the image reconstruction problem in MIAS. In addition, L-curve criterion is presented to determine the number of the latent variables in PLS. Simulations show the proposed PLS method has a better performance on image reconstructing than Tikhonov regularization method and truncated singular value decomposition method.