极化合成孔径雷达(SAR)图像地物并行分割分类研究与应用

项目名称： 极化合成孔径雷达(SAR)图像地物并行分割分类研究与应用

项目编号： No.U1204402

项目类型： 联合基金项目

立项/批准年度： 2013

项目学科： 地理学

项目作者： 薛笑荣

作者单位： 安阳师范学院

项目金额： 28万元

中文摘要： SAR多方面的优势使其在国民经济和国防建设中有着重要的应用，但SAR图像相干斑噪声的存在使其处理工作变得很复杂。另外，随着遥感技术的发展使SAR所能提供的数据信息量呈海量程度增大。然而SAR在灾害监测、预防及救灾等方面的应用则要求对获取的SAR图像数据做出快速而准确的解译。SAR图像分割分类是SAR图像解译中较为复杂的关键环节，SAR特别是极化SAR图像更完整地记录了地物目标后向散射信息，为详尽分析目标特性提供了更有力的数据支持，传统串行分割分类算法因速度慢而日益不能满足SAR的应用要求，加快分割分类速度非常重要。在高性能计算中，并行集群计算具有较高的性价比和良好的可扩展性，可满足不同规模的大型计算问题。本项目基于并行计算，研究极化SAR图像地物并行分割分类方法，通过实验和实际应用探讨并行分割分类中的数据自适应划分、动态负荷平衡等问题，优化并行性能，以最大限度地提高并行分割分类效率。

中文关键词： 合成孔径雷达；极化；图像分割分类；并行计算；

英文摘要： SAR has important applications in the national economy and national defense construction for its many advantages. But for the existence of speckle noise in SAR image, its processing becomes very complex. In addition, with the development of modern remote sensing technology, the data information that SAR can provide increases massively. However, in some fields of SAR application such as disaster monitoring, prevention and disaster relief ,etc., which require that the obtained SAR image data is processed rapidly and accurately. SAR image segmentation and classification is a complex key link in SAR image interpretation, SAR image especially polarimetric SAR image contains more complete back scattering information of ground object targets, providing more powerful data support for detailed analysis of target features. For slow speed, the traditional serial segmentation and classification algorithms are increasingly unable to meet the application requirements of SAR, improving the segmentation and classification speed is very important. In high performance computing, parallel cluster computing has higher performance price ratio and good expansibility, can meet the needs of different large-scale computing problems. Based on parallel computing, in the project, some parallel terrain segmentation classification methods of

英文关键词： synthetic aperture radar；polarimetric；image classification；parallel computing；