【作者】 刘国良；

【导师】 黄爱民；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2007， 硕士

【摘要】 合成孔径雷达(SAR)全天候、全天时的工作特点以及高分辨成像的工作能力,使得它在民用和军事领域得到了广泛应用。图像统计特征在遥感图像分析中得到广泛应用。论文首先将遥感图像常用的统计特征综合为共生矩阵、分形特征、灰度差分等六大类,利用它们的典型测度分别研究了波段、极化、入射角对SAR图像统计特征的影响。实验表明,基于不同机理的统计特征对SAR图像参数变化的敏感度不同,在不同应用目标的图像处理任务中,应当利用雷达系统参数和图像特征之间的关系,进行特征选择与提取,并综合利用多参数下的图像进行信息融合,才能得到较好的结果。在SAR图像仿真研究中,本文探索了一种融合地理信息的SAR图像仿真方法,可实现已知SAR图像的同一传感器变参数SAR图像模拟。该方法不仅利用了SAR成像几何模型,而且利用了目标区域已知SAR图像和对应的DEM数据,多信息融合生成目标区域SAR图像。实验结果证明,该方法不仅成像速度快,而且目标区域后向散射特性也得到了较好反映。最后,本文研究了一种利用多通道信息融合技术,提高SAR图像识别与分类准确率的可行方法。由于SAR图像特征与雷达系统参数间具有关联性,本文利用多视向SAR图像的信息互补,正确地识别了阴影区的地物类别,解决了单参数SAR图像分割中一个无法处理的难题。本论文的研究成果为进一步分析SAR图像的统计特征、结构特征变化规律,为SAR图像的分类与识别研究,以及基于SAR图像的匹配制导理论研究,提供了理论基础与实验平台。更多还原

【Abstract】 Synthetic aperture radar (SAR) imagery has found important applications due to its clear advantages over optical satellite imagery one of them being able to operate in various weather conditions.Statistical characteristics are very useful in analyzing remote sensing images. At first, six kinds of image’s major statistical characteristics based on different mechanism are introduced: co-occurrence matrix, the fractal characteristics, gray difference, histogram, auto-correlation function and the TAMURA. Then, the correlativity between band (or polarization and the angle of incidence) the statistical characteristics of the SAR image are researched. Experiments show that, under the changing of SAR system parameters, the sensitivities of different statistical characteristics are different. It means we should select and extract different image features in different tasks. Furthermore, the information fusion method, to take full advantage of the multi-parameter SAR image’s characteristics, will achieve good performance.To simulate SAR image, a new simulation method of information fusion has been showed in this paper. It can make an expected image of a different radar parameter by using any known SAR image and its geographic information (the DEM data of the target region). Experimental results show that this method of SAR image simulation is very fast, and the scattering properties of target region have also been better realized.Finally, a method using multi-channel information fusion technology has been proposed to improve the recognition ability and classification accuracy of SAR image. SAR image characteristics and radar system parameters are related. By using the complementary information of multiple images, the objects in shadow can be identified correctly. It is a difficult problem in the single parameter SAR image analysis.The paper will be helpful in further analysis of the statistical characteristics of SAR images, the rule of structural characteristics variation of the SAR image. And the results provide a method and a useful experimental platform for the research of scene matching, scene navigability analysis based on SAR image.更多还原