University of Twente Student Theses


PolInSAR coherence based decomposition modelling for scattering characterization of man-made and natural features

Shafai, Shahid Shuja (2019) PolInSAR coherence based decomposition modelling for scattering characterization of man-made and natural features.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:For urban planning, design assessment, management, and environmental protection, target delineation in an urban area is of great importance. One of the extensive uses of SAR is the classification of on-ground features and terrain. Synthetic Aperture Radar (SAR) is an integrated imaging technique which provides spatial characteristics of earth’s surface reflectivity in the microwave region EM spectrum. The remotely sensed area/target is characterized by the backscatter response which rendered as quantification of amplitude and phase represented by the scattering matrix. One of the ways to manifest information from complex SAR images is by employing polarimetric decompositions. This technique is capable of categorizing polarimetric signatures for different types of targets based on the scattering mechanisms they follow, enabling easy physical interpretation of the targets. Due to the presence of natural landcover, a clutter backscatter response is observed and incoherent decomposition modelling is employed for a meaningful characterization of the targets. Incoherent decomposition characterizes the scattering mechanism by decomposing the scattering matrix into an ensemble of second-order descriptors matrices which represent the scattering response of simple (Bragg’s surface model, dihedral scatterer) or canonical objects (oriented dipoles, spheroidal scatterer), manifesting an easier and comparable physical interpretation of the target. The polarimetric information is represented by an image which consists of decomposed scattering mechanisms (surface, double-bounce and volume scattering mechanism) as RGB composites. With this capability of polarimetric decomposition, it is possible to delineate urban targets due to their high double-bounce and/or direct surface backscatter response. However, urban targets which have varying shape or complex geometry or which are not orthogonal to radar line of sight are wrongly characterized as natural targets. Such targets show predominant volume scattering instead of double-bounce or surface backscattering resulting in false decomposition results. The methods available in the literature to cure this problem mainly focus on enhancing double-bounce scattering over volume scattering, neglecting the contribution of surface scattering. The assumptions and framework of incoherent decomposition models limits its capability to rightly characterized such targets which causes scattering ambiguity in decomposition results. The present work intends to solve the problem of predominant volume scattering by synthesizing the incoherent decomposition models with temporal phase correlation in terms of polarimetric interferometric (PolInSAR) coherence. In this work, a hybrid decomposition model is proposed which uses PolInSAR coherence and decomposition parameters (entropy and alpha angle) to characterize natural and man-made targets which face the problem of scattering ambiguity. The proposed model has been tested on RADARSAT2 PolInSAR data acquired over Dehradun region, Uttarakhand India. The proposed model can characterize man-made and natural features efficiently. There is a significant decrease in volume scattering from urban targets. The model is successfully able to determine the actual underlying scattering mechanism: either surface or double-bounce. The decomposition results have been compared with generic and modified decomposition models. The results reveal that the proposed model has out-performed existing decomposition models in terms of urban target characterization. To check the efficiency of the proposed model, it has been tested of multi-frequency SAR data of TerraSAR-X and ALOS PALSAR sensor for the same region. Also, the model has been tested on airborne sensor-based E-SAR PolInSAR data of Oberpfaffenhofen, Germany region. The results acquired from the decomposition of multi-frequency data highlight the robustness of the proposed model for characterizing urban targets and solving the problem of scattering ambiguity. Further, for airborne data, the model has outperformed compared decomposition models.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page