非平稳三重马尔可夫场优化结构模型的SAR图像多类分割新方法研究

项目名称： 非平稳三重马尔可夫场优化结构模型的SAR图像多类分割新方法研究

项目编号： No.61272281

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 吴艳

作者单位： 西安电子科技大学

项目金额： 80万元

中文摘要： 三重马尔可夫场(TMF)模型非常适合非平稳、非高斯图像的分析。本项目以研究TMF中优化结构模型的信息提取为重点，依据含斑SAR 图像的统计性质，建立非平稳TMF的SAR图像多类分割新理论框架，以综合分析、研究并建立基于非平稳各向异性模型U场划分的条件TMF的SAR图像多类分割的理论和方法；基于稀疏编码理论势能函数构建的高维TMF的SAR图像多类分割的理论和方法；基于软决策目标函数构建的模糊TMF的SAR图像多类分割的理论和方法；基于多尺度非线性各向异性扩散的快速TMF的SAR图像多类分割的理论和方法；突破SAR图像非平稳特性的描述、SAR图像复杂纹理的精确建模、SAR图像信息的全面捕获以及SAR图像稳健高效的多类分割等技术难点。本项目的研究将丰富和完善SAR 图像多类分割的理论和方法，具有重要的学术意义，并为军用和民用中SAR 图像多类分割提供全新理论和有效的新方法，具有重要的应用前景。

中文关键词： SAR图像多类分割；SAR图像空间相关性；高维TMF；条件TMF；快速TMF

英文摘要： Triplet Markov random field (TMF) model recently proposed is suitable for dealing with non-stationary and non-Gaussion images analysis. This fund focuses on the research of optimized structural information extraction from synthetic aperture radar (SAR) images, based on triplet Markov fields (TMF) model. According to the statistical properties of SAR images corrupted by speckles, the research objective of this fund is to construct novel theoretical frameworks based on non-stationary TMF to perform multi-class segmentation of SAR images, thus resolving the technical bottlenecks such as accurate model for the statistics of the non-stationary and complicated textural structures of SAR images, the global and local information extraction of SAR images and the statistical segmentation inference. This fund concentrates on four research streams: 1) the study and construction of multi-class segmentation methodology based on conditional TMF model, in which the field U is to be partitioned according to the non-stationary and anisotropic statistical model of SAR images. 2) the study and construction of multi-class segmentation methodology based on higher order TMF model, in which the higher order potentials are to be constructed based on sparse coding theory. 3) the study and construction of multi-class segmentation methodol

英文关键词： Multiclass Segmentation of SAR Images；Non-stationary property；Higher order TMF；Conditional TMF；Fast TMF