【作者】 伏思华；

【导师】 于起峰；

【作者基本信息】 国防科学技术大学 ， 航空宇航科学与技术， 2006， 博士

【摘要】 电子散斑干涉测量技术(Electronic Speckle Pattern Interferometry, ESPI)和合成孔径雷达干涉测量技术(Synthetic Aperture Radar Interferometry, InSAR)是非常相似又互相区别的两种技术。ESPI是一种对光学粗糙表面进行无损全场测量的技术,被广泛地应用于振动、位移、应变和医学诊断等各方面的测量中。InSAR是以合成孔径雷达复数据提取的相位信息为信息源获取地表的三维信息和变化信息的一项技术。它们都以干涉条纹为主要研究对象,并且通过对干涉条纹图的处理获取最后要测量的物理量。但无论是ESPI还是InSAR,其干涉条纹图中都存在比较严重的散斑噪声和去相关噪声,给它们的应用带来了很大的困难。本文分别对ESPI及InSAR干涉条纹图处理技术进行了深入研究,提出了系列等值线相关干涉(Contoured Correlation Interferometry,CCI)方法。创新点主要有以下四点:(1)提出了ESPI干涉条纹图的CCI生成方法。针对已有条纹图生成方法的主要弱点—乘性噪声严重,本文在大量实验及理论分析的基础上,提出了ESPI干涉条纹图的CCI生成方法。由于选取了最优的条纹等值线窗口,能生成高质量的干涉条纹图,几乎不含有任何散斑噪声。(2)提出了基于CCI条纹图的ESPI相位提取方法。利用CCI条纹图两个突出优点:不含有乘性散斑噪声和条纹正则性好,提出基于CCI条纹图的相位提取方法,包括条纹中心线法、从单幅条纹图提取相位场的方法、单步相移提取相位场的方法。(3)提出了三幅子图(Three Parts)的InSAR复图像对配准方法。该方法只需从复图像对的四个实虚部数据(四幅子图)里任取三个(三幅子图)就可以完成整个配准过程。由于精配准过程的相关运算窗口为条纹等值线窗口,配准精度亦得到了提高。(4)提出了InSAR干涉相位图的CCI生成方法。针对现有共轭相乘方法存在的缺陷,提出InSAR干涉相位图的CCI生成方法。该方法只需从复图像对的四个实虚部数据里任取三个就可以生成高质量的干涉条纹图。以上(3)和(4)一起构成了完整的三幅子图的InSAR数据处理方法,为InSAR干涉相位图处理提供了一种全新的途径。其主要优点有两个:①只用到InSAR复图像对四个实虚部中的任意三个。若SAR成像过程在星上完成,则只需要将3/4的数据传回地面,即可以实现InSAR后续数据处理过程,从而大量节约传输时间并减少对带宽的占用。②由于运算窗口为条纹等值线窗口,配准精度比原方法有提高,配准后复图像对的相干性增强。生成干涉相位图中的去相关噪声得到了有效的抑制,并且尽可能保持相位信息不受损害。这些都为下一步的相位解缠及数字高程的高精度提取创造了有利的条件。除了以上四个比较重要的创新点,本文的主要工作还体现在以下两个方面:(1)对条纹方向图的求取方法进行了深入的分析,详细介绍了本文中使用的两种条纹方向图的求取方法,对这两种方法的处理精度用模拟条纹做了精度分析与对比,为InSAR及ESPI条纹方向求取及计算窗口尺寸的选择提供了重要依据。(2)ESPI相位主值图滤波是相位解缠前必须进行的重要工作。中值滤波及均值滤波不宜直接用于相位主值图的去噪处理,它们会严重损害主值图的跳变信息。本文针对ESPI相位主值图解缠面临的主要困难,提出了将条纹等值线滤波方法与sine/cosine滤波方法相结合应用于ESPI相位主值图降噪处理的方法,取得了比较理想的滤波结果。更多还原

【Abstract】 Electronic Speckle Pattern Interferometry (ESPI), and Interferometric Synthetic Aperture Radar (InSAR), are two closely related techniques. The former is a whole-field non-destructive technique to measure the optical rough surface, and it is widely used in the measuring of vibration, displacement, strain and medical diagnosis; the latter finds its application in measuring the topography of a surface, its changes over time, and changes of the characteristics of a surface. Both of them take interferogram as their main object of research, and both obtain the last physical values for measurement through interferomgram processing. However, either ESPI or InSAR encounters great difficulties in their applications due to the serious speckle noise and the decorrelation noise existing in the interferograms.This dissertation is based on a thorough exploration into interferograms processing techniques and proposes a method named Contoured Correlation Interferometry (CCI) as a solution to the problem mentioned. Its originality is observed in the following 4 aspects:(1) CCI Method to Generate ESPI InterferogramsIn order to overcome the major shortcoming of the existing fringe patterns generation methods, i.e., the inevitable accompanying serious multiplication noise, this dissertation puts forward a CCI method to generate ESPI fringe patterns. Because this method takes the optimal fringe-contoured window, the fringe patterns generated are of high quality and almost void of any speckle noise.(2) ESPI Phase Retrieval Methods Based on CCI Fringe PatternsIt takes advantage of the two prominences of the CCI fringe patterns: a. Being free of multiplication speckle noise; b. Good normalization, to propose several phase retrieval methods based on CCI fringe patterns, i.e., fringe centerline method, single-fringe-pattern phase field extraction method, and single-phase-step phase field extraction method.(3) Co-registration Based on Three Parts of Two Complex Images for InSARIt only needs three arbitrary parts of the two complex images instead of four parts which are necessary for the existing co-registration methods. Furthermore, the co-registration precision is improved due to the fringe-contoured window used.(4) CCI Method to Generate InSAR Phase ImagesIt also takes only three parts of the two complex images instead of four for the conventional method to generate high-quality phase images. It proves an efficient tool that reduces speckle noise while preserving the phase derived, which solves one of most difficult problems in InSAR data processing.Among the above mentioned points, (3) and (4) together piece up a full picture of InSAR data processing method using only three arbitrary parts of the two complex images, which provides a totally novel approach to conduct InSAR data processing. It has two major advantages:①It requires only three arbitrary parts among the four parts of the InSAR complex image pair. Suppose the SAR image formation is fulfilled on the satellite, it asks only 3/4 of the data to be transmitted to the earth for the consequent InSAR data processing and thus to save transmission time to a large extent.②Coherence of the co-registered complex image pair gets enhanced because the fringe-contoured window is adopted as computation window and the co-registration precision is increased compared with the previous practice. Consequently, the decorrelation noise of the phase images generated is suppressed effectively and it is made possible that the phase information stays intact. All these are favorable to the phase unwrapping and the highly precise Digital Elevation Model (DEM) retrieval which come next.Apart from the forgoing four aspects, there are other two important jobs done in this dissertation:(1) An intense analysis is made on the methods to obtain the fringe orientation map, and two of them (which are adopted in this research) are given a close introduction. Furthermore, a precision analysis using simulated fringes is carried out to make a comparison of the processing precision of the two methods, which could serve as an important reference for the InSAR and ESPI fringe orientation computation and window size determination.(2) In ESPI, the saw-tooth phase map obtained by the phase-shifting technique is inherently full of speckle noise. The high-level noise of the map must be suppressed before it is unwrapped. In accordance with the feature of the saw-tooth phase maps, an adaptive filter is developed by combining the classical sine/cosine filter and the fringe orientation information of the saw-tooth phase map. Compared with existing filters, it has a better performance on phase jump information preservation and produces no blurring effect on the phase distribution provided the filtering is implemented on the equal-phase window. Moreover, its capability of noise reduction is more powerful.更多还原