【作者】 罗环敏；

【导师】 李小文；

【作者基本信息】 电子科技大学 ， 检测技术与自动化装置， 2011， 博士

【摘要】 森林在全球水文、生态、碳循环及气候变化中起着重要作用,森林的类别、林分结构、高度及生物量等参数是林业资源信息调查中的主要参数。随着遥感技术的发展,具有极化和干涉技术优点的极化干涉SAR技术(Polarimetric SAR Interferometry,简称POLInSAR),以其独特的全天候、低成本的优势,逐渐成为森林资源调查中一种不可替代的技术,在森林制图,特别是森林参数的定量反演中,发挥着越来越重要的作用。基于极化干涉SAR的森林分类、森林高度反演模型、生物量估计模型的研究是极化干涉SAR森林遥感研究的核心问题,对其开展研究有着重要的意义。本文在理解极化干涉SAR理论的基础上,研究了基于极化干涉SAR的森林结构信息提取模型与方法,主要内容如下:1、组合极化技术和干涉技术各自的优势,利用极化干涉相干优化技术,并考虑到森林场景异质性的特点,引入模糊技术对森林进行了非监督分类。该方法基于极化散射机制分割出森林区,然后用极化干涉相干优化技术提取的极化干涉特征参数A1和A2对森林区进行初始划分,并将结果做为Fuzzy C–Mea(n简称FCM)算法的初始类别,对POLInSAR数据进行迭代分类。这样,根据数据本身的特点确定初始类别,解决了FCM算法初始值不易确定的问题,同时也避免了参数A1和A2分割中武断的边界划分问题,并引入了适合自然场景处理的模糊技术,分类性能较好。2、提出了一种高性能的森林高度反演方法。详细地研究了随机体散射(Random Volume,简称RV)、数字高程模型(Digital Elevation Model,简称DEM)和随机体-地表散射(Random Volume over Ground,简称RVoG)等森林高度反演模型,调查了极化干涉相干优化和非体去相干对各反演模型性能的改善程度,并基于RVoG通过数值仿真方式分析了衰减系数和地体散射比对极化干涉相干的影响,在此基础上,发展了一种基于地体散射比、极化干涉相干优化和非体去相干补偿的相干相位-幅度综合反演方法。和其他方法相比,该方法具有较好的鲁棒性和较高的反演精度。3、基于极化相干层析(Polarization Coherence Tomography,简称PCT)技术,深入研究了影响森林垂直方向雷达相对反射率函数的因素,发现了平均相对反射率函数曲线的特征参数对生物量很敏感。用数据仿真方式分析了地体散射比和衰减系数对反射率函数的影响,然后基于森林场景极化干涉矢量测量的仿真,讨论了极化、森林密度和类型对相对反射率函数的影响,进而利用德国宇航局机载合成孔径雷达(Synthetic Aperture Radar ,简称SAR)系统(E-SAR)获取的L波段极化干涉SAR数据,分析了不同地上生物量(Above Ground Biomass,简称AGB)水平的典型林分的相对反射率函数曲线,发现林分平均相对反射率函数曲线的形状特性和森林AGB密切相关。4、创建了一种高精度的森林AGB估算模型。利用林分平均相对反射率函数曲线,定义了9个特征参数,通过逐步回归分析法,构建了森林AGB估算模型。对模型进行了评价,由于该模型充分利用了森林垂直结构信息,具有较高的估算精度。更多还原

【Abstract】 Forest plays an important role as a natural resource in global hydrology, ecology, climate, carbon (biomass) storage and carbon dynamic cycles. The main parameters in survey of forestry resources are forest categories, forest structure, height, biomass, and so on. With the development of remote sensing technology, polarimetric SAR interferometry (POLInSAR) technology which is based on the coherent combination of radar interferometry and polarimetry, has become an irreplaceable technology for survey of forestry resources because of its unique all-weather and low-cost advantages. In forest mapping, particularly quantitative retrieval of forest parameters, it plays an increasing important role. The researches on forest classification, retrieval model of forest height and biomass estimation model using POLInSAR are essential in the studies of SAR remote sensing of forest and are of great significance for application in forest. On the basis of understanding of POLInSAR theory, models and method for extracting forest structure information by POLInSAR are studied in detail in the paper. The main results are as follows:1. Based on the complementary information contained in polarimetric and interferometric SAR data, unsupervised classification approach of forest is studied in detail by using fuzzy clustering and polarimetric interferometric coherence optimization technique. The proposed method employs scattering mechanisms to indentify forest area from polarimetric SAR data. Then the forest area is further segmented by parameters A1 and A2 obtained by polarimetric interferometric coherence optimization algorithm. A robust unsupervised fuzzy C means (FCM) classifier initialized with the results of the segmentation is applied to the polarimetric interferometric coherency data sets corresponding to the forest area. As the initial categories are defined by the characteristics of the data itself, this not only solves the problem that the initial value of FCM algorithm is difficult to identify, but also avoid the fact that the A1/A2 zone boundaries were determined in somewhat arbitrary ways. So the proposed method has good performance.2. An inversion method of forest height with good performance is proposed. Several available forest height inversion models such as random volume (RV), digital elevation model (DEM), random volume over ground (RVoG), and so on are studied in detail. Then this paper investigates that to what extent do interferometric coherence optimization in radar polarimetry and non-volumetric scattering decorrelation improve the performance of forest height inversion methods. The effects of the extinction coefficient and ground-to-volume scattering ratio in RVoG model on polarimetric interferometric coherence are analyzed by means of numerical simulation. On this basis, an integrated inversion method, which combines coherence phase with coherence amplitude information and includes polarization coherence optimization, ground-to-volume scattering ratio and compensation of non-volume scattering decorrelation, is proposed and discussed. The results show that the method is robust and accurate.3. Based on polarization coherence tomography (PCT) technique, the factors possibly affecting the radar relative reflectivity function are investigated in detail and the result is found out that the characteristic parameters extracted from the average relative reflectance functions are sensitive to the biomass. The effects of the extinction coefficient and ground-to-volume scattering ratio in RVoG model on relative reflectivity function are analyzed by means of numerical simulation. Then by applying PCT to L-band POLInSAR simulations of forest scene, the effects of the polarization, forest type and density on relative reflectivity function through extinction coefficient and ground-to-volume scattering ratio are discussed. Furthermore, based on repeated pass DLR E-SAR L-band airborne POLInSAR data, relative reflectivity function curve of different levels of typical stand AGB are analyzed. Then, it is concluded that the shape features of the relative reflectivity function curve are closely related to forest AGB.4. A forest AGB estimation model with good accuracy is constructed. Based on forest stand average relative reflectivity function curve, the nine characteristic parameters are defined and used to construct forest AGB estimation model by multiple linear stepwise regression analysis method. The model is evaluated and the forest AGB estimation accuracy is good because of the stand vertical structure information considered comprehensively in the AGB estimation model.更多还原