【摘要】：合成孔徑雷達(SAR)以其能穿透云霧、煙塵及大面積獲取地表信息和全天候全天時的特點越來越受到重視。全極化SAR (PoISAR)作為當前最先進的SAR系統(tǒng),較單極化所蘊含的信息量豐富,且能揭示和描述目標內在的散射機理。由于全極化SAR系統(tǒng)特有的成像方式,使得獲取影像單一像素的散射實際由多種散射類型疊加而成。以極化目標分解為代表的基于模型的非相干分解模型將散射矩陣分解為散射模型線性加權和的形式,能夠直觀地反映該過程,且分解出的散射分量對不同應用中物理參數的分析或反演具有重要意義,是當前研究的熱點。盡管如此,基于模型的極化目標分解存在分解不一致現象,主要表現為分解后的散射分量功率為負值,嚴重阻礙了地物的分析與解譯。本文在充分理解模型分解的基礎上,分別對三分量和四分量分解中的不一致性問題進行了詳細地探究,并發(fā)展了相應的解決方法,最后將模型分解的極化特征應用于建筑物的分割中。本文的主要研究成果如下： 1.基于全極化協(xié)方差矩陣詳細推導分析了三分量模型分解不一致產生的原因,發(fā)展了一種綜合去取向理論和廣義體散射模型的三分量極化目標分解模型,首先在分解前對協(xié)方差矩陣進行去取向處理以削弱交義極化分量；然后采用一種廣義體散射模型以適應林地中HH分量和VV分量的比值變化；最后采用歸一化處理以完全消除負功率問題,該處理能夠保持地物主導散射類型不變。實驗采用德國Oberpfaffenhofen地區(qū)L波段機載E-SAR數據進行,并與其他改進算法模型進行比較,表明綜合處理后的分解模型不僅能夠消除不一致問題,且能夠有效地增強城區(qū)中的由二面角產生的偶次散射分量。 2.基于全極化協(xié)方差矩陣詳細推導分析了四分量模型分解不一致產生的原因,為解決該問題,提出了一種綜合四分量分解模型。首先,本文提出了一個新的復混合極化相關系數分別對地物目標進行選擇性去取向和螺旋體散射功率抑制。之后,為解決四分量體散射模型單一的問題,仍采用廣義體散射模型來替代原體散射模型。最后,為完全消除分解負功率像素,在歸一化處理方法的功率限制基礎上進一步改進,該處理方法包含類似Krogager相干分解的非相干分解。通過與其他改進方法的比較,使用德國Oberpfaffenhofen地區(qū)機載L波段E-SAR數據驗證了該綜合四分量分解模型的有效性,并使用德國普拉特靈地區(qū)的全極化星載X波段的TerraSAR數據探討了對短波全極化SAR影像的適用性。 3.針對傳統(tǒng)全極化SAR影像分割的缺陷,并融合模型分解后的偶次散射分量提出了一種綜合多種極化特征和幾何特征的全極化SAR建筑物分割模型,該模型通過改進分形網絡演化算法和使用多元線性回歸模型,構建綜合多特征的建筑物分割模型,通過與傳統(tǒng)H-Alpha-Wishart等分割方法進行比較,驗證了該綜合分割模型的有效性。
[Abstract]:Synthetic Aperture Radar (SAR) has been paid more and more attention because of its ability to penetrate clouds, smoke and dust, to obtain surface information in a large area and to obtain all-weather and all-weather. As the most advanced SAR system, fully polarized SAR (PoISAR) is rich in information than unipolarization, and can reveal and describe the inherent scattering mechanism of the target. Because of the unique imaging mode of the fully polarized SAR system, the scattering of a single pixel of the image is actually superimposed by a variety of scattering types. The model-based incoherent decomposition model represented by polarimetric target decomposition decomposes the scattering matrix into the form of linear weighted sum of scattering model, which can reflect the process intuitively, and the decomposed scattering component is of great significance to the analysis or inversion of physical parameters in different applications, which is the focus of current research. However, there is a phenomenon of inconsistent decomposition of polarized targets based on the model, which is mainly manifested in that the power of scattering components after decomposition is negative, which seriously hinders the analysis and interpretation of ground objects. In this paper, on the basis of fully understanding the model decomposition, the inconsistency in three-component decomposition and four-component decomposition is discussed in detail, and the corresponding solutions are developed. finally, the polarization characteristics of model decomposition are applied to the segmentation of buildings. The main research results of this paper are as follows: 1. Based on the full polarization covariance matrix, the causes of the inconsistency in the decomposition of the three-component model are derived and analyzed in detail, and a three-component polarimetric target decomposition model based on the deorientation theory and the generalized body scattering model is developed. firstly, the covariance matrix is deoriented to weaken the intersecting polarization component, and then a generalized volume scattering model is used to adapt to the ratio change of the HH component and the VV component in the forest land. Finally, the normalization process is used to completely eliminate the negative power problem, which can keep the dominant scattering type of the ground object unchanged. The experiment is carried out by using L-band airborne E-SAR data in Oberpfaffenhofen area of Germany and compared with other improved algorithm models. It is shown that the decomposition model after comprehensive processing can not only eliminate the inconsistency problem, but also effectively enhance the even scattering components generated by dihedral angles in urban areas. two銆，

本文編號：2512230