【作者】 张振华；

【导师】 保铮；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2007， 博士

【摘要】 分布式小卫星是近年来提出的一种新概念、新体制,它利用一颗在轨星载SAR(也可以重新发射)作信号源,在其前/后发射若干小卫星接收地面回波信号,从而实现一颗卫星不可能完成的功能,如:高分辨宽幅成像,地面高程测量,地面运动目标检测,测洋流等。为了防止碰撞,小卫星和主星的间距一般比较远,所以对小卫星接收的回波数据不能直接用单基成像算法进行成像。另一方面,SAR图像又是实现分布式小卫星系统上述各项功能的前提和基础。因此双基SAR成像算法的研究成为现在雷达界的热点,也正是本文选题的一个主要原因。虽然本文是以分布式小卫星为基础,但为了突出双基SAR的特点,对机载双基SAR也进行了研究,而且主要构形为具有方位平移不变性的平行轨道双基SAR系统,因为在合成孔径时间内,收发平台都可以认为以恒定的速度等速运行。双基SAR系统的点目标斜距历程为收发斜距和的双根号形式(称为“平顶双曲线”),与单基SAR系统相比,随方位时间与最近距离的变化减弱,因此双基SAR的聚焦深度增加,徙动量减小,这是一个有利因素。但由于相位历程为双根号形式,用驻相点法很难得到其精确的多普勒域解析表达式。这对设计快速的双基频域成像算法造成一定的困难。本文提出一种基于瞬时频率法推导双基SAR回波信号二维频域解析表达式的方法。从得到的表达式,一方面可以清晰地看出双基与等效单基的关系,易于在单基SAR的基础上理解双基SAR,另一方面可以方便地分析各个参数对成像的影响。表达式中包含几个双基独有的参数:最近距离和、最近距离差、半类双基角。如果以最近距离和作为距离变量,则双基成像算法首先要解决后两个变量在方位(指多普勒)和距离(指最近距离和)的二维空变问题。通过两次泰勒级数近似可得到半类双基角和最近距离差随多普勒和距离二维变化的近似表达式,进而提出一种计算复杂度不高,且适用于高分辨宽幅的双基距离徙动成像算法(RMA)。在大斜视情况下,半类双基角和最近距离差的二维空变比较严重,用上述近似处理还存在误差。因此本文提出用数值计算方法获得这两个参数的方法,然后代入点目标频域响应函数得到一种精确的数值RMA成像算法。由于RMA算法需要进行“类stolt”插值,运算量比较大,另外,插值误差还会影响SAR图像的初相,增大后续干涉处理的相位误差。因此基于上述点目标频域响应函数提出一种不需要插值、高保相、高效率的双基CS成像算法。实现此算法的关键是把表达式中随纵向距离空变的参数进行线性化处理。上面两种成像算法的思想是把单基SAR里相应的成像算法进行推广,另一种思路是把双基的回波数据经过补偿变为某个等效单基的回波数据,然后按照单基的成像算法去成像,即等效单基法。此方法的好处是避开双根号积分问题。原理上,即使SAR具有很高的分辨率,一个像素仍相当于一复杂目标,其回波有很强的方向敏感性,双基回波不能用单基来等效。但在BORN一阶近似的前提条件下(即忽略二次和多次散射现象),可以把散射点看作是独立的点目标。这时,就可以把双基SAR的回波数据等效为一定视角的单基SAR回波数据,因为这时它们的区别只是在波程差上。常规的等效相位中心法(EPC)是把双基等距线椭圆补偿为圆心在收发中点,与椭圆相交的一系列圆。在远场条件满足时,此方法是完全精确的。在基线较长时,需要补偿掉随距离和多普勒二维变化的一个相位,并且基于如下近似:等效单基和双基的多普勒相等。此近似只在基线较短时成立。DMO方法是从地震成像学中推广过来的,此方法把双基等距线(椭圆)等效为一系列与之相切的圆,这些圆的圆心和半径(分别对应等效单基的位置和斜距)各不相同,圆半径和椭圆长轴之间的差就是需要补偿的斜距。根据SAR系统具有包络缓变和相位快变的特点,对补偿函数分两步进行:首先在距离多普勒域对影响相位的部分精确补偿,然后在二维频率域用场景中心线的包络补偿函数对全场景进行统一补偿,也可以分距离块进行更精确的补偿。本文给出用几何关系推导DMO补偿函数的全新方法,并提出了实现精确补偿的变标傅氏变换法,但此方法的缺点是运算量很大。DMO等效单基改变了双基回波信号的距离向波数,从而带来很大的好处,首先,它没改变双基回波的多普勒波数;其次,它没有改变双基系统的方位和距离分辨率特性;另外,它没有改变双基SAR图像的干涉特性(此单基和双基得到的SAR图像完全相干),完全可用于基于SAR图像的干涉后处理。分布式小卫星SAR系统(比如Cartwheel构形的主辅式小卫星SAR系统)的一个重要应用就是利用多个接收小卫星得到的双基SAR图像进行地面高程测量和动目标检测,完成这两项任务首先要保证每幅SAR图像具有高保相性(即在成像处理过程中没有破坏初相),其次还要保证图像对之间具有较高的相干性。本文分析了双基SAR图像的初相,以及地形起伏对双基SAR图像的影响,在DMO等效单基思想的基础上,提出一种利用两幅双基SAR图像的等效视角谱匹配来提高具有水平基线双基SAR的干涉预滤波方法。此方法可以把双基SAR图像对中由于基线引起的非相干成分全部滤除。信噪比问题是分布式小卫星SAR系统面临的一大难题,在主星带伴随小卫星群体制的分布式小卫星系统中,主星是已有的SAR卫星,用来发射信号,其发射信号的平均功率比较低,一般为100-300瓦,这一功率是按照自发自收模式下满足最小功率孔径积的要求设计的,一般不会留有很多的余量。而接收回波信号的形佬堑奶煜呙婊冉闲?方位向直径一般设计为1-2米,(Cartwheel中天线面积为4.5m~2),因此小卫星分布式SAR系统的信噪比会比主星低,我们分析了信噪比下降后对提高成像分辨率、干涉处理等方面的影响,并提出用多通道相干或非相干处理提高信噪比的方法。更多还原

【Abstract】 This dissertation addresses topics in one main area of radar imaging formation: bi-and multistatic synthetic aperture radar (SAR) image formation, especially effective bistatic SAR focusing algorithms.The concept of distributed radar satellite constellations has been proposed as a new remote sensing technique in the literature and currently under research. It is suggested that one on-orbit radar satellite can be used as an active transmitter while several other micro-satellites act as passive receivers, and the active transmitter, if necessary, can be a new launched radar satellite. This kind of spaceborne implementation offers extra advantages compared to the monostatic scheme, namely high resolution wide swath imaging, topographic imaging or DEM generation, ground moving target indication (GMTI) and current measurement. It is clear that the bistatic radar constellation is the basis of the distributed spaceborne system, and thus the bistatic SAR is the fundamental research topics. Generally, bistatic SAR imaging is more involved than the monostatic SAR case due to the considerable separation of the transmitter and the receiver for the sake of avoiding potential collision of platforms. Therefore, the bistatic SAR data can not be processed sufficiently by using the monostatic focusing algorithms, i.e., the conventional monostatic image formation algorithms are not well suited to this type of bistatic constellation. This dissertation mainly discusses the spaceborne cases, but in order to illustrate the bistatic SAR characteristics the airborne bistatic cases are also discussed. The bistatic SAR systems can be considered as Azimuth-Invariant during an aperture time, provided that the bistatic transmitter and receiver move in a constant velocity in a parallel constellation. This case is referred to as parallel bistatic SAR system for simplicity. This dissertation focuses on this kind of systems.Due to the two square rooted terms in the slant range history of the bistatic SAR, it is very difficult, if not impossible, to obtain the analytic expression of the phase history of echoed data in the 2-dimensional frequency domain by using the principle of stationary phase, which adds to the difficulty in developing fast focusing algorithms for bistatic SAR data.We propose an analytic formulation method in the 2-dimensional frequency or wavenumber domain based on the concept of the instantaneous Doppler wavenumber. In the formula, the space-variance of the two new defined parameters, i.e., the difference of the closest distances and the half quasi bistatic angle, are detailedly discussed.A computation effective bistatic Range Migration Algorithm (Bi-RMA) is proposed under some rational approximation to the space-variance of the above two parameters. This proposed bi-RMA can be used to process the wide swath bistatic SAR data sufficiently. Another benefit of the developed analytic expression lies in the fact that it becomes easier to understand the inherent relationship of the bistatic SAR and its equivalent monostatic SAR based on the formula, and the parameters influence on the resulted image can be explained in a quite straightforward way.Unfortunately, in the high squinted bistatic cases, the space-variance of the above two parameters may be so heavy that the bi-RMA can not be used to focus the data any longer. We derive a numerical method to handle this case. The two parameters, the difference of the closest distances and the half quasi bistatic angle, are calculated by using the numerical approach, respectively. Using the developed analytic expression of the echoed signal in the 2-dimensional frequency domain in combination with the obtained parameters, we get the accurate bi-RMA in the sense of numerical computation.However, the computational load of the proposed bi-RMA is relatively heavy yet due to the quasi-Stolt interpolation used in the processing. Furthermore, the interpolation is not a phase preserving operation both in the bistatic and monostatic image formation, which is not expected for the possible interferometric applications. Also on the basis of the developed frequency domain analytic expression of bistatic data, we propose a phase preserving, computation effective bistatic Chirp Scaling (bi-CS) algorithm. The linearization of the range-variant parameters is one of the most important steps.The above proposed two image formation methods are both based on the analytic expression in the 2-dimensional frequency domain and the conventional monostatic SAR focusing scheme, and the bistatic transfer function is derived to handle the data. Another type of method is to compensate the bistatic data into the equivalent monostatic data and then processed using the monostatic algorithms. Thus, the troublesome two square rooted terms problem is successfully pushed away. However, it should be make clear that one resolution cell is composite of many individual scatterers even with a very high resolution capability, and the echo data is so sensitive to the radar-target vector that, strictly speaking, the bistatic data can not be fully represented by the equivalent monostatic data. But if Born approximation is followed, it is feasible. Now the bistatic data is identical to the monostatic data at the certain squint angle since the difference between them lies in the scaling factor of radial wavenumber.In the conventional EPC (equivalent phase center) processing method, the bistatic iso-range ellipses are compensated into the iso-range circles of the equivalent monostatic one with the origin located in the middle of the baseline. It is accurate if the far-field assumption is satisfied. Actually, the compensation depends on the range lines of the imaging scene and the Doppler frequency of the EPC monostatic SAR. Due to the fact that the bistatic SAR and the EPC monostatic SAR are not identical in the Doppler frequency and we obtain the bistatic Doppler domain data by directly using FFT over the raw data, one assumption is adopted in the EPC compensation, i.e., they are identical in the Doppler frequency. Thus, the EPC method can only be used to the bistatic data from a short baseline system.The so-called DMO (Dip Move Out) technique use the SMILE concept of seismic processing as a reference. In DMO bistatic image formation algorithm, the bistatic iso-range curves (ellipse) are compensated into a series of monostatic iso-range curves (circle) which are all inscribed circles of the former ellipse. But both the centers and the radius of the circles are not identical to each other, respectively. Each sample of a bistatic snap shot can be considered as a series of monostatic different range samples along the bistatic baseline after the slant range differences are compensated respectively. Suppose that the SAR echoed signal has the property of slow fluctuation in envelop and fast fluctuation in phase, the compensation can be divided into two main steps: first, the phase differences can be fully removed in the range Doppler domain. Second, the bulk envelop difference can be compensated using compensation function of the swath center line in the 2-dimensional frequency domain, and of course the more accurate result can be obtained by using a blocking compensation in range dimension. A scaling Fourier transform approach is developed, which is proved to have the capability of accurate compensation but the computational load is too high to use.The DMO approach has several advantages in implementation. Although the radial wavenumber is changed, the azimuth Doppler of the bistatic echoes remains not distorted. Second, both the range and azimuth resolution are not changed. The last, but most important, is that the interferometric phase is not distorted, thus the bistatic interferometry is possible.Two main applications of the distributed micro-satellites radar systems, for example, the Cartwheel constellation, are the topographic imaging (finally as DEM) and GMTI by using the multi-channel SAR data from different receivers. For the purpose of interferometry processing, both high phase preserving of the image formation processing and the high coherence of the multi-channel SAR image pairs are expected. The fixed phase of the bistatic echo is analyzed as well as the affect over bistatic image caused by the terrain relief. On the basis of the DMO concept, we propose a coherence improvement pre-filtering method for the bistatic SAR system with an along track baseline. Using this method, the baseline-induced non-coherent components of the bistatic image pair can be fully removed.SNR is a challenging problem for distributed micro-satellite radars. Suppose one master satellite in a constellation, the average power of the transmitted signal is generally low, namely 1 to 3 hundred Watts. Due to the relative small antenna aperture size of the slave micro-satellites, for instance, 1-2 meters in azimuth, their received power will be quite low compare with the master satellite. On the basis of the analysis of this reduced SNR affects on the resolution and interferometric applications, we approach this problem by using the proposed coherent and non-coherent techniques over the multi-channel received echoes, respectively.更多还原