Research On High-Precision Imaging Methods For Synthetic Aperture Radar

Synthetic aperture radar (SAR) is a kind of all-weather and all-day microwave imaging system. The human makes much higher request for SAR 2-D resolutions to obtain more abundant and more detailed target scattering information and to improve the abilities of target detection, recognition and perception. High-resolution bandwidth synthesizing and motion compensation are the keys to break through the range resolution and azimuth resolution, respectively. In the aspect of bandwidth synthesizing,the compensation for space variant range error in one cluster is the main bottleneck for the improvement of synthetic range profile performance. In the aspect of motion compensation, how to estimate the motion error efficiently and precisely under the condition of complex trajectory is the key point for the improvement of azimuth focusing performance. To improve the 2-D imaging precision for SAR, this dissertation focuses research on synthetic bandwidth method and motion compensation method under the support of the National Special Fund for High Resolution Earth Observation.The main contents and innovations are as follows:1. Image-domain synthetic bandwidth methods based on back projectionIn stepped-frequency SAR, the different transmission and receiving positions of sub-band pulses will result in space variant slant range error, which will deteriorate the synthetic range profile performance. The conventional synthetic bandwidth methods compensate the range error all by the reference target point, and thus they do not accommodate the space variance of range error. For this problem, an image-domain synthetic bandwidth method based on time-domain back projection is proposed. In this method, the range error of every pixel is compensated simultaneously when the azimuth compression is operated. This method overcomes the difficulty brought from the space variance of range error and maintains the continuity of phase on the edge of sub-band spatial spectrum. As a result, the sidelobes of synthetic range profile are well suppressed.To improve the efficiency of synthetic bandwidth method in image domain, a synthetic bandwidth based on frequency-domain BP is then proposed. With the help of frequency-domain BP, the concept of spatial spectrum of SAR image is elucidated,which reveals the essence of bandwidth synthesizing in SAR image space. Then the relationship between the cutting positions of sub-band spatial spectrum and the system parameters, with which the automation for spectrum cutting is realized and the synthetic image performance is persevered at the same time.2. Synthetic bandwidth methods embedded in back projectionFor the contradiction between high range resolution and wide swath in the bistatic stepped-frequency SAR that works in the sequential transmission mode, the system of bistatic multichannel stepped-frequency SAR is demonstrated to reduce the requirement for PRF (Pulse Repetition Frequency). This system has the ability of synthesizing high-resolution range profile while maintaining wide swath. Although the image-domain synthetic bandwidth method based on back projection can compensate the space variant range error precisely, but the coherence of range resolution directions for all pixels should be required, and thus they are not applicable in bistatic multichannel stepped-frequency SAR, in which the range resolution direction is space variant. For this problem, a synthetic bandwidth method embedded in back projection is proposed.In this method, the space variant range error compensation and bandwidth synthesizing are embedded in the process of BP imaging and the range error compensation and wideband spectrum reconstruction for each pulse and each pixel are precisely implemented. This method achieves high-precision range profile reconstruction under the condition of space variant range resolution direction.3. High-efficiency and high-precision autofocus back projection algorithmsIn the aspect of motion compensation under the condition of complex trajectory,the global autofocus BP algorithm works well and has wide applications. However, the requirements for memory and computation in global autofocus BP are both very large.For this problem, a local autofocus back projection algorithm is proposed. The rationality of focusing the whole scene just through auto focus for local region is mainly analyzed. Besides, how to select the local region for phase error estimation is also analyzed. Compared with global autofocus BP, the efficiency of autofocus BP is largely improved while focusing performance is persevered. For image formation of wide scene,where the space variance of motion error is very prominent, the global and local autofocus methods loss efficacy. For this problem, an antenna phase error estimation method is proposed, in which the position errors are modeled. After detail derivation for the gradient of objective function, the conjugate gradient method is used to solve the optimizing model. The results show that the focusing performance of this method is largely improved when the motion error is space variant.4. Compensation for large dynamic error and high-precision 2-D spectrumThe above autofocus BP algorithms will loss efficacy when the variation range of initial positioning error is very large. For the large dynamic error compensation, a novel autofocus BP algorithm is proposed after analyzing the influence of 3-D initial position errors on SAR imaging. The genetic algorithm is applied to solve the optimizing model.To improve the efficiency of genetic algorithm, the two-step strategy of 'coarse estimation + precise estimation' is further proposed. This method realizes meter-level precise and stable estimation for initial altitude error and finally removes the geometry distortion effect effectively. Besides, in the aspect of frequency-domain imaging model for high-speed diving SAR, the three-order slant range model and two-dimensional spectrum model are demonstrated through Chebyshev optimal approximation.Compared with four-order Taylor spectrum model, the new concise spectrum model provides reference for simplifying the high-precision frequency-domain imaging algorithm.