Physical and statistical based decomposition of Polarimetric Synthetic Aperture Radar images of Arctic Sea ice

The studies about the climatic changes have always more underlined the importance of the climatic balance of the Arctic regions. For this reason the need of monitoring the Arctic becomes always more urgent. To measure the sea ice thickness, the sea ice cover, the motion of the glaciers and to discri...

Full description

Bibliographic Details
Main Author: Arienzo, Alberto
Format: Master Thesis
Language:English
Published: UiT Norges arktiske universitet 2015
Subjects:
Online Access:https://hdl.handle.net/10037/11108
Description
Summary:The studies about the climatic changes have always more underlined the importance of the climatic balance of the Arctic regions. For this reason the need of monitoring the Arctic becomes always more urgent. To measure the sea ice thickness, the sea ice cover, the motion of the glaciers and to discriminate the various kind of ice are only several of the challenges about the Arctic monitoring. But the extreme climatic conditions make the Arctic one of the most inaccessible regions on the Earth. Radar imaging and in particular polarimetric radar imaging provide indispensable instruments in this challenge. In our thesis work we analyzed a common topic in radar polarimetry: the model-based decompositions. Such decompositions have the goal of interpreting the scattering mechanism for each single pixel in the polarimetric image through statistical instruments, as the covariance or coherency matrix, and physical instruments, as the main laws of the electromagnetism in the context of the scattering theory. The model-based decompositions are typically characterized by a large number of unknowns, the parameters of the target, but usually they cannot be estimated for lack of enough equations. Typically, the model-based decomposition problems are underdetermined and in order to find an unique solution it is necessary fixing some parameters or making some prior assumptions. The ideal condition would be to have more equations in order to uniquely resolve the system, without approximations. This it is exactly the goal of our thesis work, introducing new equations using the fourth-order moments. Investigating such a possibility we analyzed a particular specific model-based decomposition for the sea ice: the Sea Ice Two-Component decomposition. The simulations have been made using test pattern especially built in such a way to have a solid and effective reference of the quality of the decomposition. Only after we tried with the real sea ice image of the Fram Strait, Greenland. The obtained test pattern results have shown a significant improvement in the parameters estimation compared to the second-order case. As regards the real simulations, we cannot affirm the same thing of the test patterns. However, we think the good test pattern results, are a preliminary confirm of the usefulness of the fourth-order moments in the model-based decompositions. To use the additional equations given by the fourth-order moments, it has enabled us: 1) to find an algebraic solution without fixing any parameters, 2) the possibility of including the product model and so to get the information texture for any model-based decomposition. However, often to find an analytic solution is very complicated. For this reason, we implemented an optimisation algorithm with a relative normalisation strategy that it allowed us: 1) to retrieve a solution when an algebraic solution cannot be found and retrieving a larger number of free parameters in respect to the traditional model-based decomposition, 2) to obtain smooth image thanks to the speckle robustness of the optimisation algorithm. Concluding, our work shows a preliminary possibility of using the fourth-order moments in the model-based decompositions.