基于时频分析的极化SAR图像人造目标提取技术研究

发布时间：2018-05-03 14:51

本文选题：极化SAR + 人造目标提取　；参考：《国防科学技术大学》2014年博士论文

【摘要】：本文围绕“极化SAR图像人造目标提取”课题进行研究,基于时频分析技术挖掘目标及其背景杂波的散射特性,提取时频特征增强目标抑制杂波,并融合其它有效极化特性,构建出提取效果好、抗干扰能力强、适用范围广、贴近工程应用的无监督目标提取算法。实验过程针对多个复杂场景进行分析,重点关注复杂背景条件下建筑物、舰船两类人造目标提取工作。论文主要内容以及创新点如下:1.复杂电磁干扰有可能影响SAR成像质量,对于严重影响图像质量的显性干扰,由于干扰噪声强度不低于信号强度,导致任何后续解译工作无法进行,本文介绍了该类干扰抑制处理流程,并利用受干扰的海冰场景数据进行方法验证;对于不影响图像幅度信息,但有损相位信息的隐形干扰,当对数据进行时频分析或者极化特性分析时,该类干扰会影响后续特征提取。为此,本文设计了干扰抑制方案,实验证明了干扰抑制有利于后续时频特征的准确提取。2.基于时频平稳特性分析,提出了时频平稳性度量方法,借助于人造目标的时频非平稳特征以及自然背景地物杂波的平稳特征,对两类复杂场景进行了时频分析。第一,在复杂城区场景中对建筑物目标进行提取,验证方法的有效性,比较利用单极化以及全极化数据的实验结果;第二,针对极地复杂海冰场景,分析海冰独特的散射特性,利用时频平稳度描述子对舰船和海冰背景进行鉴别处理,增强了两者对比度,有利于准确提取舰船目标。3.基于时频相干特性分析,设计了时频相干性度量方法,并提出了新的基于时频相干度量的极化特性描述子,分析了三类复杂场景:第一,对于复杂城区场景,相干度量方法能够有效提取建筑物,基于新的极化特性描述子对提取出的建筑物进行极化特性分析,能够有效地对建筑物高度等相关信息进行估计反演;基于双波段的极化SAR数据,比较分析不同波段条件下相干散射子的提取结果,实验分析结果与理论估计相吻合。第二,对于复杂虚影场景,将不同时频分解模式对虚影的抑制程度进行比较,同样对该场景下舰船目标进行极化特性分析。实验证明时频相干度描述子能够对舰船进行有效地提取,并极大减少了虚影可能引起的虚警。第三,对于复杂海冰场景,利用时频相干度描述子对舰船提取、检测以及鉴别,并利用所提出的时频极化特性描述子刻画舰船极化特征,分析结果符合目标真实结构。4.基于多特征融合方法,构建了多特征融合的人造目标提取方案,具体内容如下:第一,将时频平稳性与相干性结合起来,定性分析两者关系,基于融合特征对建筑物进行最大程度的提取,取得了满意结果。第二,探讨了时频特征与极化干涉相干度特征融合方法,有利于进一步对城区环境进行地物分类。第三,面向低分辨率星载城区数据,融合时频相干度与极化反射对称等多项极化特征,非时频特征弥补了时频分析降低分辨率的不足,时频特征有着独特抗干扰性能,两者优势互补,融合策略能有效提取建筑物目标。第四,鉴于城区的高度不均匀性,提出了新的基于非高斯G0模型的时频平稳度描述子,并融合圆极化相关系数进行阈值设定,对不同朝向建筑物进行提取,效果良好。总的来说,论文提出的一系列时频分析方法以及特征描述子,能够有效地对极化SAR人造目标进行提取,基于多类复杂场景不同传感器数据进行了性能检验评估,可以达到工程应用水平,为进一步目标解译研究奠定了基础。
[Abstract]:This paper focuses on the subject of "polarizing SAR image artificial target extraction". Based on time frequency analysis technology, it excavate the scattering characteristics of target and background clutter, extract time frequency characteristics to enhance target suppression clutter, and fuse other effective polarization characteristics, and build a good extraction effect, strong anti-interference ability, wide application range and close to engineering application. Unsupervised target extraction algorithm. The experimental process is aimed at a number of complex scenes, focusing on the complex background conditions of buildings and ship two types of artificial target extraction. The main contents and innovation points are as follows: 1. complex electromagnetic interference may affect the quality of SAR imaging, and the dominant interference which seriously affects the quality of the image is caused by the complex electromagnetic interference. The interference noise intensity is not less than the signal intensity, resulting in any subsequent interpretation work that can not be carried out. This paper introduces this kind of interference suppression processing flow, and uses the interfered sea ice scene data to verify the method; for the stealth interference that does not affect the image amplitude information, but the phase information is damaged, the time frequency analysis of the data is performed or When the polarization characteristics are analyzed, this kind of interference will affect the subsequent feature extraction. Therefore, the interference suppression scheme is designed in this paper. The experiment proves that the interference suppression is beneficial to the accurate extraction of the time frequency characteristics of the subsequent.2., and the time frequency stability measurement method is proposed, with the aid of the time-frequency nonstationary feature and self - frequency characteristics of the artificial target. However, the time-frequency analysis of two kinds of complex scenes is carried out for the stationary features of the background ground clutter. First, to extract the building targets in the complex urban scene, verify the effectiveness of the method and compare the experimental results of the single polarization and the fully polarized data. Second, the unique scattering characteristics of the sea ice are analyzed for the extremely complex sea ice scenes. The time frequency stationary descriptor is used to identify the ship and sea ice background, which enhances the contrast between the two. It is beneficial to accurately extract the ship target.3. based on the time-frequency coherence analysis. The time frequency coherence measurement method is designed, and a new polarimetric characteristic descriptor based on time frequency coherence measurement is proposed, and three kinds of complex scenes are analyzed. First, for the complex urban scene, the coherent measurement method can effectively extract the building. Based on the new polarization characteristic descriptor to analyze the polarization characteristics of the extracted building, it can effectively estimate the building height and other related information. Based on the dual band polarization SAR data, it compares and analyzes the phase under different wavelengths. The results of the dry scatterer are in agreement with the theoretical estimation. Second, for the complex virtual shadow scene, the suppression degree of the virtual shadow is compared with the different time-frequency decomposition modes, and the polarization characteristics of the ship target are analyzed in the same scene. The experiment shows that the frequency phase dry descriptor can effectively extract the ship. Third, for complex sea ice scenes, the time frequency coherence descriptor is used to extract, detect and identify ships, and the characteristics of ship polarization are described by the time frequency polarization descriptor. The analysis results conform to the real structure of the target.4. based on multi feature fusion method, and the multi feature is constructed. The main contents are as follows: first, combining time frequency stability and coherence, qualitative analysis of the two relations, based on fusion characteristics to the maximum degree of extraction of buildings, obtained satisfactory results. Second, the time frequency characteristics and polarization interference coherence feature fusion method is discussed, which is beneficial to the entry. Step to the urban environment classification. Third, facing the low resolution spaceborne city data, fusion time frequency coherence and polarization reflection symmetry, such as multi polarization characteristics, non time-frequency characteristics make up for the time frequency analysis to reduce the resolution, the time frequency characteristics have unique anti-interference performance, the two advantages complement each other, fusion strategy can effectively extract building. Fourth, in view of the high heterogeneity of the city, a new time frequency stationary descriptor based on the non Gauss G0 model is proposed, and the threshold setting is combined with the correlation coefficient of circular polarization to extract different orientations. In general, a series of time frequency analysis methods and feature descriptors proposed in this paper can be used. It is effective to extract the polarimetric SAR artificial target and evaluate the performance of different sensor data based on multiple complex scenes. It can reach the level of engineering application and lay a foundation for further target interpretation research.

【学位授予单位】：国防科学技术大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】