Detection Of Terrain Surface Changes Using PolSAR Images And Multi-PolSAR Simulation Of Forest With Snow Cover

In recent decades, Synthetic Aperture Radar (SAR) technology has been constantly developed, and its applications are becoming more and more wide and deep. With the emergence of high-resolution SAR, single-and multi-(full-) polarization SAR images have good applications in terrain change detection and forest monitoring fields. This paper researches the properties of statistical and probability of SAR images, Markov random field theory, the methods of change detection in target and terrain change detection before and after the earthquake automatically, theory and properties of the target scattering, full-polarization SAR image decomposition, radar polarimetry, Stokes vector and the Vector Radiative Transfer (VRT) theory, the full-polarization Mueller matrix's simulation of the forest with snow cover and H-αparameter characteristics.In terrain change detection aspect, a method of two thresholds expectation maximum and Markov random field (2EM-MRF) is presented to automatic detection of terrain surface changes after Wenchuan earthquake, May 12th 2008 using multi-temporal ALOS PALSAR images.As an example in Beichuan area, three kinds of terrain surface changes, i.e. scattering enhanced, reduced and no-changed, are automatically detected and classified.Using the tool of Google Earth, the surface change situation after the earthquake can be demonstrated in multi-azimuth views as animated cartoon.The detection and classification are also compared with the optical photographs.The study shows that multi-temporal SAR technology with high-resolution and full polarimetry and fusion with other remote sensing data are essentially effective to monitoring natural disasters.Our approach is feasible to adopt for future development.In the simulation of Mueller matrix aspect, suitable forest model and snow cover model must be selected.The forest model is composed of a layer of discrete random different types of non-spherical particles, including the type of trunk, main branch, often sprawling and leaves.The snow cover model is the dense Rayleigh particle model.In the forest and snow cover models, simulate the backscattering Mueller matrices of both the normal forest and the forest with snow cover cases when the incident waves are L-band and X-band.Solve the coherency matrics from the obtained Mueller matrice,.Then, calculate the H and a parameters. At last, analyze the change of H and a parameters both the normal forest and forest with snow cover models.During the graduate three years, two aspects have been researched. One is to use the synthetic aperture radar images in the same area before and after the earthquake, through the 2EM-MRF method to detect the terrain changes and analysis the changes's types and causes.Another is the simulation of Mueller matrices of the forest and forest with snow cover models using the VRT theory and then calculating H, a parameters, analyzing the impact of snow cover. Meanwhile,with the fast developing technology, many things need to be researched deeply and to be completed.