| Abstract:Lardge scale inventories of vegetation resources can provide fundamental data to analyze and quantify the dynamic process of ecosystem, carbon cycle and climate. Special emphasizes had been put on maping vegetation and extracting vegetation vertical structures. Maping of vegetation and vegetation vertical structures are of very importance for topography surveying and inventory of forest resource. With the development of geographic national conditions monitoring, the requirements of geographic fundamental informations, such as mapping of vegetation and extraction of vegetation vertical structures, are imperative. Monitoring measures of high accuracy and large scale are urgent to be developed. PolInSAR provides an opportunity to resolve the urgent requirements. For vegetation area, PolInSAR treats the vegetation as volume rather than plan under the sight of passive optical remote sensing. This advanced strategy makes PolInSAR a unique observatory, especially for vegetation volume, to derive maps and vertical structures of vegetaions. In addtioan, based on the theoretical basis of PolInSAR, models and methods of vegetation classification and vertical structure extraction are dicussed in this paper. The main research contents are as follow.(1)A PolSAR vegetation classification based on the characteristics of vegetation morphology is proposed in this thesis. Anisotropy and degree of orientation randomness of scatterers are used to express the characteristics of vegetation morphology. In this way, broad-leaved forest and coniferous forest can be classified effectively. Firstly, we make an intensive study of decomposition models of Freeman and Yamaguchi, and analyze their principles and disadvantages. This help us to deeply understand the principle of Neumann. Based on as mentioned above, the classification principle is proposed by using the anisotropy and degree of orientation randomness. After that, cluster analysis is conducted by Wishart distribution. Lastly, E-SAR and SIR-C/X-SAR data of Oberpfaffenhofen, Germany, are selected for validation of new method. Compared with Wishart-Freeman and Wishart-Yamaguchi, the new method is more sensitive to the the characteristics of vegetation morphology, which makes broad-leaved forest and coniferous forest can be classified effectively. (2) A PolInS AR vegetation height extraction method using complex least squares is proposed in this thesis. The new method resolve the limitation that the conventional methods donnot pay much attentions to priori statistical errors and redundant observations. The conventional methods donnot pay much attentions to priori statistical errors and redundant observations. Embarks from the modern theory of adjustment, the RVoG model is expressed by complex adjustment. The methods of function model linearization, adjustment criterion, stochastic model and parameter estimation are constructed. After that, complex least squares adjustment vegetation extraction methods based on RVoG、 RVoG+VTD and three-layer vegetation scattering model. Lastly, the methods are tested with airborne and spaceborne SAR data. The new method can not only take account of the redundant observations well, but also can ruduce the ill-posed problem caused by poor poor geometrical dispersion. The results show that the proposed method is superior to existing method in terms of the accuracy.(3) A PCT method based on complex least squares is proposed. The priori statistical errors is considered and Wiener-SVD is employed to overcome the ill-posed problem of design matrix. The priori statistical errors are always ignored in conventional PCT method which also does not own robust strategy to deal with the ill-posed problem. After the analysis of PCT principle, a PCT expressed via complex least squares is founded, and Wiener-SVD is adapted to overcome the ill-posed problem. BioSAR2008data are selected for validation of the modified method. The results shows that the new method is more robust than the conventional method. |
PDF Full Text Request