【作者】 李玥；

【导师】 李亚超；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2017， 硕士

【摘要】 随着合成孔径雷达(SAR)成像技术的发展和完善,基于雷达图像的匹配技术已成为研究热点并广泛应用于众多领域。SAR图像匹配技术是将两幅来自不同成像时间段、不同成像角度及不同传感器的SAR图像进行相似度量的一种技术。如何高效地提取两幅SAR图像间的稳定信息,并做相似性配准是SAR图像匹配技术的研究重点。SAR图像匹配作为SAR图像处理的基础核心技术,其配准精度及稳定性是SAR图像处理技术的奠基石。因此,国内外图像处理研究专家致力于提高图像匹配的正确性及有效性。本文深入研究了SAR图像匹配技术。首先针对SAR图像特性,分析SIFT匹配算法在SAR图像匹配领域的适用性;从成像角度分析SAR图像与光学图像的辐射及几何特性差异,并根据差异性分析SIFT算法在SAR图像匹配应用中的局限性。分析表明,SAR图像的斑点噪声对SIFT匹配算法影响最大。其次,考虑到SAR图像斑点噪声对SIFT算法在SAR图像匹配应用中的影响,本文采用适用于SAR图像斑点噪声的SAR-SIFT算法,分析算法对SAR图像斑点噪声的抑制作用,并从算法计算复杂度及提取特征点的稳定性等方面进行改进,提出一种基于多极值抑制改进拉普拉斯变换的SAR-SIFT算法。改进后SAR-SIFT算法提取的特征点在图像上分布更加均匀、稳定且精度达亚像素级。在不影响算法匹配精度情况下,改进后的SAR-SIFT算法运算效率更高。最后,研究在大仿射形变情况下的SAR图像匹配算法。SAR图像的成像时间及成像视角的不同造成SAR图像间具有较大仿射形变。SIFT及SAR-SIFT等尺度不变特征算法对仿射形变较大的SAR图像不能发挥良好的匹配性能。本文将ASIFT算法与改进的SAR-SIFT算法相结合,运用于大仿射形变的SAR图像匹配中,最终的仿真实验结果表明,新的ASIFT算法具有良好的匹配性能。更多还原

【Abstract】 Along with the development and improvement of synthetic aperture radar(SAR)imaging technology,image matching based on SAR image has been a new hotspot in the field of SAR image processing recently.SAR image matching was the technology of finding some similar measures in two SAR images taken from different periods or different imaging angle or different sensors.How to efficiently extract the stable information in two images and match them with effective methods was the key of the image matching research.As the core technology of SAR image processing,the accuracy and stability of image matching were the cornerstones of all.Therefore,image processing researchers abroad bent themselves to improving the correctness and effectiveness of the image matching.This dissertation mainly focused on the SAR image maching,comprehensively analyzed the applicability of the SIFT algrithm based on SAR image.SAR imaging theory and optimal imaging was described for analysing the radiation and geometric feature differences between the two types of images.The application limitations of the SIFT algorithm in SAR image matching were disscused based on the differences.The simulation experiments proved that the speckle noise in SAR image had the most significant impact on the SIFT algorithm applied in SAR image matching.In consideration of the application limitations of the SIFT algorithm in SAR image matching,this dissertation studied SAR-SIFT algorithm which was put forward by Flora Dellinger and adapted to the SAR image speckle noise,investigated the applicability of SAR-SIFT algorithm to speckle noise in SAR images,meanwhile,improved the SAR-SIFT algorithm to enhance the speed and the stability of the feature points.An improved SAR-SIFT algorithm based on multi extremum suppression and Laplasse transform was proposed in this dissertation.The simulation experiments showed that the improved SAR-SIFT algrithm could be more uniform and stable in feature points which reached an accuracy of subpixel level,and the improved algorithm could be more efficient.The method of matching two SAR images under different view conditons of affine deformation was discussed.An optimal matching technology named Affine-SIFT algorithm with affine invariant including the affine transformation,image perspective sampling,implementation and the effectiveness was studied.A new SAR image matching method conbined the improved SAR-SIFT with ASIFT algorithm was proposed.The final simulation results showed that the new algorithm was much better.更多还原