Study On Fast Time Domain SAR Imaging And Three Dimensional SAR Motion Compensation Methods

As is widely used in the national defense and civil industry, synthetic aperture radar(SAR) is a microwave remote-sensing equipment that is capable of working under all weather conditions day and night, in a long distance with high 2 dimensional(2D) resolution or wide swath. Owing to the development of technology, SAR system is no longer working only in narrow-beam sidelooking stripmap mode with linear track, but also in much more challenging situations like none-linear track, extremely long synthetic aperture, super high resolution, highly squinted, wide swath and multiple beam-steering modes. Back-projection algorithm(BPA) accumulates energy in time domain pulse by pulse. It could be properly used in all kinds of trajectory and all imaging modes theoretically. But it suffers from the heavy burden of massive computation. By dividing the synthetic aperture into pieces and constructing coarse images individually, then fusing images in local polar coordinates(LPCs) stage by stage, the fast factorized back-projection algorithm(FFBPA) obtains a well-focused full-resolution image gradually. The amount of computation is greatly reduced comparing with that of BPA. However, huge amount of 2D image domain interpolations are still required in image fusion stages to improve accuracy. The accuracy and the efficiency of FFBPA are restricted by the interpolation method. In this dissertation, the theory of FFBPA is analyzed and improved, and three faster time-domain imaging algorithms are proposed, which can also be utilized in many different imaging modes mentioned above.The imaging results obtained by single-channel or along-track multi-channel SAR systems are the projections from the scenario in 3D to the range-azimuth plane. The Multiple-inputmultiple-output(MIMO) SAR system with a cross-track linear array has the ability of resolving along the third dimension. Thus it could provide more information about some special terrain, such as an active volcano. 3D imaging has aroused general interests in recent years. Small airplanes, especially the unmanned aerial vehicles(UAVs), equipped with the linear-array MIMO SAR systems(such as the ARTINO from Germany) are severely disturbed by airflow. Given no high precision inertial navigation system(INS) equipped, a good motion compensation(MOCO) method is necessary. The 3D motion error model of the linear-array MIMO SAR is analyzed in this dissertation. And a new algorithm based on joint multi-channel auto-focusing technology is proposed which can not only compensate the translation error but also the rotation error with high accuracy.The main content of the dissertation is summarized as follows.1. It is known that the interpolation number of FFBPA is up to the length of subapertures. The fundamental reason of it lies in the adoption of low efficient BPA in the first processing stage. Hence, the shorter the first stage subapertures are, the faster FFBPA will be. However, shorter subapertures mean more image fusion stages. Too many 2D image interpolations are utilized causing degrades in image quality. A new fast algorithm combining PFA and image fusion is proposed in this dissertation. The efficiency of the new algorithm rises as the length of the subapertures. Meanwhile, less interpolations result in high precision. The computation burden and the memory requirement of the new algorithm are analyzed. The applications of the new algorithm in many SAR modes are also discussed. Finally the validity of the new algorithm is testified by simulation and real data experiments.2. Aiming at the contradiction between the computation efficiency and accuracy caused by the large number of interpolations in FFBPA, a fast interpolation-free method is proposed. The offsets of an image grid projected from one LPC to another are analyzed. The image fusion is implemented using range shifting, angle rotating and linear scaling in both dimensions. No interpolation is utilized in this new method. Therefore, the new method is of higher efficiency and accuracy than the interpolation-based FFBPA. The computation loads and the applicable scope of the new method are also analyzed. Simulation and real data experiments verify the superiority of the new method over the interpolation-based FFBPA eventually.3. LPC is chosen to be the image fusion coordinates in LPC in FFBPA. The reason is that much lower sampling frequency is needed in subaperture imaging without causing aliasing in LPC. Nonetheless, the grids are different in each LPC causing distortion in images. As a result, 2D image interpolations are necessary to maintain accuracy in image fusion. Cartesian coordinates(CC) is, however, very outstanding due to its uniform sampling. Image fusion is much easier to accomplish in CC. Yet, high azimuth sampling rate is required in CC leading to massive computations. In this dissertation, a spectrum compression technique(SCT) is introduced and a new fast method named Cartesian factorized back-projection algorithm(CFBPA) is proposed. The azimuth spectrum is compressed twice in range time domain and range frequency domain respectively in CFBPA. The squinted spectrum is also corrected.The SCT makes the fast factorized BP integral in CC possible. No error is introduced from interpolation. CFBPA can achieve precision as high as BPA does theoretically. The application of CFBPA in non-linear track SAR modes is discussed in this dissertation. The computation load is also analyzed. The ability to coping with high resolution non-linear SAR data is testified through simulated 0.1m spaceborne SAR data processing experiment. The high efficiency and high precision is further verified through 0.2m airborne real-data processing experiment.4. Due to the baseline in the third dimension, the linear-array MIMO SAR system could resolve in three dimensions. Unfortunately, the imaging results could be severely disturbed by the airflow, especially for the small aircraft without INS equipped. Traditional MOCO methods only deal with translational motion error and ignore the rotational motion error. Nevertheless, the rotational motion error could be considerable for the 3D imaging in practice. To solve this problem, a new MOCO method for 3D imaging is proposed in this dissertation by mathematical modeling. The influence of the rotational motion error on the linear array is analyzed. The motion error estimation accuracy is improved using the multivalue optimization technique. The translational motion error and the rotational motion error are both compensated. Therefore, the imaging results are of higher quality. Finally, the validity of the new MOCO method is verified through simulated 3D imaging experiments.