【作者】 马仑；

【导师】 廖桂生；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2008， 博士

【摘要】 更广阔的观测区域、更高的空间分辨率和更逼真的三维地形获取是星载SAR系统永无止境的追求。传统的单星SAR系统越来越受到很多制约,包括成本、可靠性和基线难以做长等。二十世纪九十年代中期,人们提出了一种新的天基雷达体制,即利用多颗小卫星构成超稀疏雷达阵列进行编队飞行。小卫星具有重量轻、体积小,研制周期短、成本低,发射灵活等一系列优点,而且还可以形成批量生产,因此利用小卫星组成星座的成本会低于一颗传统的大卫星,而其整体功能则远远优于一颗大卫星。利用多颗小卫星编队飞行在具有单航过优点的同时,能够提供多个空时自由度,可以大大提高综合孔径雷达(SAR)的性能,包括:宽域、高分辨率SAR成像,低速地面运动目标检测,多基线、单航过干涉SAR等。然而分布式小卫星雷达具有很多优点的同时也面临着问题和挑战,例如空间阵列超稀疏分布、回波信号存在距离或/和多普勒模糊以及存在多种误差源等等,必须克服这些困难才能充分发挥分布式小卫星雷达的优势。本论文从信号处理的角度,针对小卫星发射-小卫星接收体制,主要研究解决多普勒模糊、实现宽测绘带和高分辨率SAR成像处理的数据处理方法。包括分布式小卫星的阵列构形为沿航向线阵以及三维超大立体阵列两种情况下的地面场景重构以及误差估计问题。本论文的主要工作可以总结如下:1、小卫星分布式雷达沿特定的轨道超稀疏分布,这相当于在空间中形成一个多通道采样系统。而且,由于最小天线面积条件的限制,单颗小卫星的空间采样是欠采样的。为了消除距离\多普勒模糊的影响,通过联合其它小卫星来增加空间采样率,这实际上与时域多通道采样方法的基本思想是一致的。对多通道采样与信号重构方法的研究将为解决多普勒模糊提供理论基础。沿用时域多通道采样方法将宽带信号进行多通道延时采样的思路并且结合频域多通道采样方法中频带分割的基本思想,提出了一种对通道增益误差及时延误差稳健的重构完整宽带信号的新方法。该方法在频域利用自适应波束形成技术来恢复宽带信号的完整带宽。由于采用了稳健的自适应处理技术,在各通道存在误差的情况下,该方法仍然能够稳健地恢复宽带信号的完整带宽。2、小卫星的天线孔径较小,不满足最小天线面积约束,其接收的回波信号会存在距离/多普勒模糊。因此必须展开或抑制距离/多普勒模糊才能获得无模糊、高分辨率的SAR图像。而且,相对于上述的时域多通道采样系统,小卫星分布式SAR系统将引入更多的误差源,如三大同步误差(时间同步误差、频率同步误差和波束同步误差),基线误差、通道误差和偏航(导致各个子孔径天线非沿航向直线排列)等等。本论文对分布式SAR系统中存在的各种误差源进行分析、根据它们对多普勒模糊抑制的影响进行分类并给出相应的补偿方法;并给出了一种对补偿残留误差稳健的多普勒模糊抑制方法,该方法能提高空域导向矢量对残留误差的稳健性;最后利用一组实测的多通道机载数据验证了以上方法的有效性。3、如果仅仅实现SAR以及GMTI功能,则编队小卫星沿航向直线分布(具有很短的垂直航向基线)为最佳构形。由于沿航向直线分布的阵列在地形高度方向不具有分辨能力,用于自适应处理的样本不会受到地形起伏的影响。然而,这种卫星编队构形不具有地形高程测量(InSAR)功能。分布式小卫星InSAR系统不但可以获得大观测带和高方位分辨率的二维SAR图像,而且还可以获得地形高程信息。然而,在InSAR卫星编队构形(即具有长垂直航向基线)下获取大测绘带、高分辨SAR图像面临很大的挑战。其中,地面单元的阵列导向矢量随地形高度和距离的剧烈变化对于获取足够的独立同分布样本(i.i.d)进行自适应处理带了挑战性。另一个关键挑战是宽带阵的包络配准与回波信号的采样模糊耦合在一起,使得包络配准与解决模糊变得更加困难。为解决分布式小卫星InSAR系统所面临的这些问题,提出了利用该系统重构宽测绘带、高分辨率三维地形的新方法。该方法利用高分辨SAR成像技术来获取足够多的样本,然后利用空-像域联合子空间正交投影技术获得SAR图像中所有模糊分量的高度信息,根据高度信息对所有像素中的每一个模糊分量进行包络配准,最后利用自适应波束形成技术取出所有多普勒模糊分量实现宽广地面场景的高分辨率三维重构。4、分布式SAR系统中存在多种误差源,而偏航以及波束指向误差是影响SAR图像聚焦的主要因素。如何消除以上误差的影响进行SAR聚焦是分布式SAR系统实现宽测绘、带高分辨率SAR成像处理、GMTI以及InSAR测高的基础。从图像域出发,提出了两种新的SAR自聚焦方法,分别为基于最大全变差准则和基于DCT准则的SAR图像自聚焦算法。它们分别利用信号的最大全变差以及DCT估计相位误差系数。与其它图像域自聚焦方法相比,以上两种方法计算量较小,更易实现,并且可以通过多维搜索估计任意阶次的相位误差。更多还原

【Abstract】 A new conceptual implementation of spaceborne synthetic aperture radar (SAR) system in which several small satellites fly in constellation forming a highly sparse array was presented in the mid-1990s. Small spaceborne system has many advantages, such as light weight, small size, short development time, low cost, flexible launch and so on. Further more it can batch produced. Consequently, forming a constellation of small satellites not only has far lower cost than a conventional spaceborne SAR system but also better performance.The formation flying distributed small satellites can provide multiple and long baselines in single-pass observation mode, thus greatly improving the performance of interferometric SAR (InSAR) and ground moving target indicator (GMTI). The coherent combination of several SAR images obtained from different observing angles can improve the image resolution and provide accurate geometric information. Furthermore, combining a broad illumination source with multiple small receiving antennas placed on separate formation-flying micro-satellites, we can obtain high resolution SAR images of wide areas. However, several challenging problems are also introduced by the constellation SAR regime at the same time. These problems include high sparseness of the array, range or Doppler ambiguities, many kinds of error sources and so on. These problems should be carefully considered to ensure that the advantages of the constellation can be achieved.In this doctoral dissertation, the approaches to reconstruct ground scean with wide swath and high resolution and the problems of error estimation are studied. The main returns of this doctoral dissertation are listed as follows:1. The small satellites in constellation distribute sparsely along specific orbit, which is to say that they form a multi-channle sampling system. And for a single small satellite, it is under sampling in spatial domain. Combining multiple small satellites in constellation to increase the spatial sampling rate the distributed spaceborne SAR system can restrain range or Doppler aliasing. This idea is accordant with the fundamental thought of time-domain multi-channel sampling method. The research of multi-channel sampling and its signal reconstruction method can provide the theory foundation for resolving Doppler ambiguity. Introducing the multi-channel delay sampling idea and combining the idea of frequency band segmentation in frequency multi-channel sampling method, we propose a new reconstructing method which is robust to delay error and channel gain error. The method retrieves the complete bandwidth by using adaptive beamforming thechnique in frequency domain. Due to the adoption of robust adaptive processing, the proposed method can retrieve the complete bandwidth of broad band signal robustly, even when the error exists.2. The minimum antenna area constraint cannot be satisfied by individual small satellite in constellation. Range and/or Doppler ambiguity will be introduced by small antenna inevitably. To obtain the SAR image with wide swath and high resolution, we must restrain Doppler aliasing firstly. Compared with time-domain multi-channel sampling system, distributed spaceborne SAR system will introduce more error sources, such as timer error, beam pointing error, frequency synchronization error, baseline error, channel error, yawing (it distorts the along track linear formation of the sub apertures) and so on. In this dissertation all kinds of errors existing in the real system is classified according to the impact on the performance of Doppler aliasing restraining, approaches to compensate the errors are proposed and a method of restraining Doppler aliasing which is robust to remanent errors is presented. Finally all the methods above are verified by using a set of airborne multi-channel measured data.3. Linear array along the track (or cross-track baseline is very long) is the best configuration for the distributed spaceborne SAR system, if only SAR and GMTI need to be implemented. The samples will not be affected by the terrain fluctuation in this case, because the linear array along the track does not have the ability of resolving the height of terrain. However, this type of configuration can not achieve height measurement (InSAR). Not only can constellation InSAR system achieve wide swath and high resolution two-dimensional SAR image, but also it can obtain the terrain height information. Nevertheless, highly sparse three-dimensional array brings new challenges for data processing, where the violent change of beam pattern (array steering vector) along with the height and range of the terrain make it difficult to obtain sufficient independent and identically distributed (i.i.d) samples for adaptive processing (to restrain Doppler aliasing). The envelope registration of broadband array coupling with the sampling aliasing of the echo is the other important challenge generally, which makes it more difficult to resolve Doppler ambiguity and register the envelope. To overcome these difficuties, this dissertation proposes a novel method of reconstructing wide-swath and high resolution three-dimenssional topography. The method obtains sufficient samples in high resolution SAR image domain and acquires the height information of all aliasing components in SAR image by using joint space-image subspace projection technique. According to the height information the envelope of the every aliasing component is further registered, and all aliasing components in the pixel are extracted by using adaptive beam forming technique. Finally reconstructing three-dimensional topography is finished.4. There are many error sources exist in the distributed spaceborne system, yawing and beam pointing error is the main factor which affect SAR focusing. It is the basis of wide-swath and high resolution SAR imaging, GMTI and InSAR to avoid the affection of the error sources above to focus the SAR data. Starting with complex phase-degraded SAR image, two novel SAR autofocus algorithms are presented in the dissertation. They estimate phase error coefficients by using the total viriation and DCT of the signal. Compared with other SAR autofocus algorithm based on image domain, these two methods above are of less computational complexity and easy to implement.更多还原