【作者】 王金峰；

【导师】 皮亦鸣；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2010， 博士

【摘要】 合成孔径雷达(SAR)三维成像,既具有传统SAR全天时、全天候、高分辨率成像及电磁穿透等优势,又避免了二维成像时三维场景空间投影到二维成像平面时的模糊问题,在军事侦察、地球遥感、海洋研究、环境保护以及灾情监测等方面具有广泛的应用前景,正逐渐成为雷达技术领域的研究热点之一。SAR层析三维成像技术,利用不同入射角轨迹上获得的多个二维SAR成像结果,基于层析原理在二维成像平面的法线方向上进行孔径合成,从而实现高分辨三维成像。该技术既不需要复杂的飞行轨迹控制、额外的定位精度要求,又拥有真正的三维成像能力,在雷达成像领域具有很高的研究价值。然而,SAR层析三维成像技术还不完善,在理论、系统和应用等方面都面临着问题与挑战。本文以SAR层析三维成像技术为研究对象,重点针对系统理论、图像配准、噪声抑制、稀疏轨迹、机载平台拓展等关键技术开展了较为系统、深入的研究,完成的主要工作和贡献如下:1)完善了SAR层析三维成像理论与系统。本文以层析三维成像的视角分析了雷达信号的层析成像特征,详细地阐述了SAR三维层析成像理论,推导了SAR层析三维成像的性能。分析了SAR层析三维成像系统中二维成像分辨率、轨迹定位误差、图像配准精度、轨迹数量及分布等系统因素对SAR层析三维成像的影响。研究了SAR层析三维成像系统中图像配准、成像等关键问题的解决方法。2)提出一种基于区域特征的多SAR图像配准算法。该算法通过水平集函数与配准函数的同步迭代求解泛函极值获得图像的配准函数实现图像配准。该算法抑制了SAR图像相干斑对区域特征识别的影响,并有效避免了特征识别误差向配准过程的传播。3)针对SAR层析三维成像中复杂的噪声环境,提出了一种适应复杂噪声环境的SAR层析三维成像算法。该算法基于高阶统计理论,将成像建模为噪声中的谐波恢复模型,提高了SAR层析三维成像系统对图像间去相关因素及环境的适应性。4)提出了一种稀疏轨迹SAR层析三维成像算法。该算法基于多散射中心假设,将成像问题建模为信号稀疏表示问题,再利用稀疏贝叶斯学习算法求解实现成像。降低了SAR层析三维成像对轨迹的要求,提高了SAR层析三维成像的适用性。5)初步设计了一种多天线机载SAR层析三维成像系统。该系统采用单发多收的组合模式,降低了SAR层析三维成像对机载平台的导航精度和飞行密度的要求,降低了飞行风险,拓展了SAR层析三维成像在机载平台的应用。本文重点研究了SAR层析三维成像实际应用中的几个问题,丰富了SAR层析三维成像系统和理论。随着SAR层析三维成像研究的逐渐深入,应用条件的逐渐成熟,SAR层析三维成像技术必将得到广泛的应用。更多还原

【Abstract】 In recent years, the technique of three-dimensional Synthetic Aperture Radar (SAR) imaging has been becoming one of the focuses in RADAR field because of such specific advantages as all-time, all-weather, high-resolution, electromagnetic penetration, no projection blurring and broad application prospects in military reconnaissance, earth remote sensing, marine research, environmental protection and disaster monitoring. Based on the tomography, SAR tomography (TomoSAR) can generate high resolution three-dimensional imagery by combining multiple SAR images acquired on flight paths slightly separated in the elevation direction. In the field of radar imaging, TomoSAR has important research value because it have a true three-dimensional imaging capability, does not require complex flight trajectory controls, additional positioning accuracy requirements. However, many problems and challenges are needed to be resolved for perfecting the TomoSAR.This dissertation carries out research on TomoSAR with focuses on the key techniques such as systems theory, SAR images co-registration, suppressing noise interference, imaging with sparse baselines and airborne extended application. The main work and contributions are presented as follows:1) To some extent, TomoSAR theory and system are supplemented. With the view of tomography, the characters of radar signal are analyzed firstly. Secondly, the TomoSAR three-dimensional imaging principle is presented and the performances of TomoSAR are deduced. Then the effects from the factors of system are discussed such as the resolution of SAR images, the errors in track positioning, the precision of SAR image co-registration, noise, the number and distribution of tracks et al. Fourthly, the solutions to some key problems are studied and presented.2) A novel feature-based method for SAR images co-registration is proposed. In this method, the co-registration is formulated as a functional optimization problem based on level set firstly, then the level set function and the registration function are interleaving evolved iteratively to register the images. By this method, the speckle effect on feature recognition is suppressed and the errors occurred in recognition is avoided to impact registration simultaneously.3) A new imaging approach is proposed for TomoSAR based on higher-order statistics. By formulating the imaging to harmonic retrieval problems based on higher-order cummulants, this approach can deal with colored (and whiten) Gaussian noise and symmetric (and non-symmetric) non-Gaussian noise. The TomoSAR is extended to more complex situations.4) To deal with TomoSAR imaging with sparse baselines, an approach is proposed based on sparse signal representation. Firstly, the imaging is formulated as a signal sparse representation problem through the hypothesis that the signal in elevation direction are from limited scattering centers. Then the imagery is obtained by solving the representation problem through sparse Bayesian learning algorithm. With this approach, the TomoSAR can obtain imagery from more flexible SAR tracks both number and distribution.5) An airborne TomoSAR system configuration with multi-antennas is presented where all the antennas are distributed along the wings, one antenna transmits and receives electromagnetic waves, and the others are receivers only. With this configuration, the TomoSAR can work with fewer flights, less flight risk, larger flight positioning precision.In this dissertation, some practical issues of three-dimensional SAR tomography imaging are mainly discussed. To a certain extent, the work presented in this dissertation enriches the TomoSAR systems and theories. With the devolopement of TomoSAR, it will be applied in various fields.更多还原