Vegetation Height And Underlying Ground Altitude Estimation Based On Multi-Baseline POLINSAR Images

As part of the earth ecosystem, the vegetation is of fundamental importance to agricultural production and the carbon cycle. It is, however, difficult, at a global scale, to obtain forest areas and to analyze the vertical structure of vegetation using conventional botanical methods. As an active remote sensing approach, the synthetic aperture radar (SAR) provides global observation ability all the time. The subsequent advanced interferometric and polarimetric techniques (INSAR and POLINSAR) show a potential capability to vegetation researches.In radar theory, the intricate vegetation vertical structure function (VVSF) gives an inherent relation between the radar complex backscattering coefficient and the special vertical altitude. The current VVSF algorithms can be partitioned into three categories, which contain VVSF parameterization, Fourier-Legendre expansion and tomography, to solve the dozens of unknowns. The parameterization method, such Random Volume over Ground (RVoG) and Oriented Volume over Ground (OVoG) models, utilizes the empirical exponent VVSF expression to simplify the solution process by POLINSAR images. The Fourier-Legendre method uses the Fourier-Legendre polynomial series to approximate the original VVSF on POLINSAR systems. Furthermore, the tomography method (TSAR) estimates the strict VVSF directly by tens of repeating-pass INSAR images. To reduce the images number, we try to solve the special unknowns of VVSF, the vegetation height and underlying ground altitude, but not VVSF itself. In this dissertation, two major contributions are as follows:(1) The original VVSF parameterization method, RVoG model, is extended into multi-baseline observed condition. The wind-driven canopy flicker and the vertical canopy fill factor are considered in the forest interferometric process. Furthermore, the multi-baseline algorithm is utilized to solve the RVoG underlying ground phase and height-extinction ambiguous problems. In the real data experimental part, the RVoG model and the baseline influence are also evaluated by multi-baseline POLINSAR data. (2) Technically speaking, the RVoG model only uses the polarimetric information to obtain the vertical location of certain backscattering mechanisms. Actually, the underlying ground caused polarimetric orientation angle (POA) can also be extract by sophisticated polarimetric backscattering model. In this dissertation case, the novel adaptive ground scattering model is proposed to approximate the underlying mechanism. Combining with the Arii vegetation volume scattering model, the multi-baseline POLINSAR set is utilize to solve the adaptive model. To simplify the estimation procedure, the coherence matrix entropy is employed to estimate the solution rangeing of Arii volume randomness parameter. Then, a new objective optimum function is proposed to solve the unknowns.