DEVELOPMENT OF POLARIMETRIC SAR TECHNIQUES FOR URBAN APPLICATIONS

Abstract

Global urbanization stands as a prominent and imminent challenge in both the present and future. Extensive evidence suggests that they have far-reaching consequences for various natural and human systems, including those pertaining to energy, water, food, biodiversity, climate, and human health. The global urban transition signifies that a majority of the world’s population currently resides in urban areas. However, there is a dearth of accurate global data regarding the whereabouts and spread of human settlements in urban regions. There is still much to learn about the true extent of this issue. Earth observation (EO) imagery is an efficient method for addressing the absence of precise data regarding the layout and spatiotemporal growth of human settlements. Therefore, in urban remote sensing, the classification of human settlements is an important topic. Researchers working on earth observation applications have come in favor of monitoring urbanization through the land cover categorization. To achieve this, Synthetic Aperture Radar has proven to be highly effective due to its advantages which include round-the-clock acquisition, all-weather functionality, and sensitivity to targets’ geometric and physical characteristics. SAR systems featuring fullpolarimetric configurations have yielded particularly encouraging results. The true vector nature of electromagnetic radiation is best preserved in full or quad polarimetric mode, which makes this mode suitable for the comprehensive backscatter study. The data obtained by a full-polarization configuration enables the creation of the full-polarization scattering matrix, providing comprehensive polarimetric scattering details of the target. Major space agencies have developed sophisticated polarimetric radar sensors for earth observation due to a growing interest in radar polarimetry applications. Some agencies have already launched their sensors, while others are in the process of developing and launching new ones. A few of the novel fully polarimetric sensors include ALOS-PALSAR (L-Band, JAXA - Japan), TERRASAR (X-Band, DLR - Germany), RADARSAT-2 (C-Band, CSA - Canada), UAVSAR by (L-Band, NASA /JPL - United States), and NISAR (L- and S-Band, jointly by NASA-ISRO) scheduled for launch soon. Interpreting the data obtained from these sensors constitutes the foundational stage of any information extraction/ processing methodology. Using target decomposition theorems, the measured values from full polarimetric data are converted into the parameters that correspond to the physical significance of the scatterer. These theorems explicate the scattering mechanisms by leveraging information from the full polarimetric data. The basis of land-cover categorization is the accurate interpretation of the underlying scattering mechanism since different land-cover categories entail distinct forms of scattering. One direct approach to achieve this is by decomposing full polarimetric data into a linear combination of physical scattering mechanisms, popularly known as "model-based decomposition". A considerable number of model-based decomposition approaches have been documented in the open literature, and researchers are keen on the development of additional methods. Rooted in the advancement of model-based decomposition techniques, this thesis aims to study urbanization through Polarimetric SAR datasets. To accomplish this, a three-stage framework has been proposed. The approach begins with the development of algorithms for the extraction of areas of human settlements. These areas usually signify dense heterogeneous regions consisting of diverse man-made and natural features. The second approach involves the identification of these features by the development of an adaptive edge detection technique. In the last phase, the height of buildings present in the residential urban areas is computed, and outcomes are validated with the ground truth data. Once validated, it is then correlated with the census data to determine the urban population growth. The aforementioned three stages constitute three working chapters of the thesis. The suggested strategies outlined in these chapters contribute to the existing body of literature by introducing three novel approaches. The performance of these approaches has been qualitatively and quantitatively analyzed against the conventional and state-of-the-art methods. An accurate land cover classification technique is an essential prerequisite to performing the remote sensing of urban areas. Therefore, Chapter 2 begins with a thorough analysis of challenges and advancements in the model-based decomposition technique. The development and shortcomings of the PolSAR edge detection techniques are covered in the latter portion of Chapter 2, which is then followed by a discussion on the mapping and monitoring of urbanization. In Chapter 3, two entirly different model-based decomposition techniques have been proposed for the extraction of urban areas. Among these two developed approaches, the first one has been employed in Chapter 4 for the development of an adaptive edge detection technique. While the second approach is implemented in Chapter 5 as the preliminary step in the assessment of urbanization.