APPLICATION OF RADAR WAVE POLARIZATION FOR THE STUDY OF GROUND TERRAIN PARAMETERS

Abstract

Polarization diversity in radar systems has opened a realm of new applications for imaging radar systems that promises advancement in both the theoretical modeling and the experimental characterization ofthe radar scattering behavior ofthe objects. Recent launches of fully polarimetric space-borne SAR (Synthetic aperture radar) sensors are boon towards the highly improved global imaging and mapping of the terrestrial covers. This in turn requires the exploration of the existing methods and developing the newer ones for the purpose. Futuristic SAR missions are more focused towards deploying bistatic configuration, which in spite of being subject to complex processing endow with more flexibility as compared to the conventional monostatic missions. Bistatic SAR systems offer considerable degrees of freedom in choosing transmitter and receiver trajectories that provide additional information observations for the extraction of scene and target parameters. Experimental observations are the primary access to support the conceptual studies of new remote sensing methods and thus it naturally becomes the pressing need for the employment of bistatic geometry of radar systems for various applications. Taking an incentive of above SAR designs and their surpassing applications, the present research work is divided in two parts. First part is devoted to the investigation and critical analysis of the methodologies for landcover classification using polarimetric data and the second part constitutes the characterization of surface parameters (i.e., surface roughness and soil moisture) for facilitating their efficient retrievalwith bistaticmodeof the radar. Research work is organized into seven chapters. First chapter provides a basic platform of the research work by presenting a brief introduction, motivation and problem formulation along with the details of the experimental and the satellite data sets. Some of the basic concepts of polarimetry relevant for the work have been also discussed in the chapter. A brief literature review of the related works is presented in chapter II. Chapter III of the thesis explores the capability of polarimetric PALSAR (Phased Array type L-band Synthetic aperture radar) data for landcover classification. In this chapter, analysis of polarization signatures in linear and circular polarization for certain targets has been carried in out which lays the basis for fusion of linear and circular polarizations for classification. The main focus of this work is to highlight the effect of circular polarization along with linear polarization on land cover classification. Linear polarization is confined to a single plane containing the direction of propagation while for circular polarization; wave radiation is in horizontal, vertical and planes between them [162]. Thus, circular polarization is found more effective in examining the targets having different orientations or alignment with respect to radar line of sight (LOS). Taking this in account the information about the target from both linear and circular polarization can be merged to enhance the classification results. Thus, in this chapter, initial linear basis has been converted to circular one(left/right handed circular). Subsequently, polarization signatures have been plotted for few targets in both the bases and the information gathered from each case has been compared [125]. Taking the ground of this information, feature sets have been made on combining linear basis and circular basis components together. Minimum distance classifier has been used to classify the image into three broad land covers i.e., water, urban and agriculture. The analysis of the results led to the conclusion that employing circular polarization with linear polarization can enhance the overall accuracy and classification accuracy ofcertain classes of interest. Chapter IV presents a comparative study of different target decomposition techniques for validating their suitability in classification of land covers with the help of few generic supervised classifiers. A number of coherent and incoherent techniques exist for the decomposition of targets based on scattering mechanisms. Except for few works, the need for comparing the existing decomposition techniques for their suitability in terrain classification remains an open area for research. Incoherent decomposition techniques are more suitable for describing natural targets but had been less reported. Since, the choice of a suitable classification method for particular decomposition technique affects the classification accuracy of various land covers. Therefore, in this chapter a detailed critical analysis has been carried out to check the individual performance of each classification method on the selected decomposition techniques for major land cover classification. Eigen vector based H/A/alpha method and scattering model based decompositions have been used [62, 212, 216, 218]. An effort has been made to compare the classification results obtained from the listed decomposition methods. For the purpose, algorithms have been applied on the processed PALSAR data. Classification of the decomposed images from each of the methods has been done using four supervised classifiers (parallelepiped, minimum distance,

Mahalanobis and maximum likelihood) [84, 169]. Ground truth data generated with the help of ground survey points, topographic sheet and Google earth has been used for the computation of classification accuracy. Parallelepiped classifier gave better classification accuracy of water class for all the decomposition methods excluding H/A/Alpha. Minimum distance/ Mahalanobis distance classifier gave better classification results for urban class. Maximum likelihood classifier performed well as compared to other classifiers for classification of agriculture class. In chapter V, a novel effort has been made for the characterization of surface roughness parameters i.e., vertical roughness (rms height), 's' and horizontal roughness (i.e., correlation length), 7' at X-band for a dry bare sandy soil in bistatic specular direction. There is a dearth of experimental investigations carried in bistatic mode using correlation length as the roughness parameter along with rms height. Thus, the main aim of this work is to focus on the importance of inclusion of correlation length as a roughness parameter alongwith rms height using ground-based bistatic scatterometer data. Such an experimental investigation thus paves the path for the effective modeling of roughness parameters that can be used for the futuristic bistatic configurations. For the accomplishment of this task, observations have been carried out in bistatic specular direction and some of the existing models have been used for retrieving the roughness parameters. Stationary phase approximation (SPA model), a theoretical model has been modified for the bistatic specular case for retrieving the roughness parameters. A simplistic incidence angle approach developed initially for emissivity [179] has been reformulated for bistatic scattering coefficient and used effectively for the retrieval of surface parameters. An indoor experiment under controlled conditions was performed and periodic rough surfaces were created by varying roughness in horizontal and vertical directions. Experimental observations were taken over nine soil fields with s (0.39-0.73 cm) and / (1.00-2.32 cm) varied under controlled conditions for different incidence angles and like polarizations. A critical analysis of the angular response of scattering coefficient with both roughness parameters has been done for both like polarizations (HH- and W polarizations). It was noticed that W polarization provided higher dynamic range of a° at different roughness than HH-polarization. Incidence angles less than 50° showed better results for observing the effect of roughness parameters on a°, however 40° was found as the best suitable incidence angle at a frequency of 10 GHz for observing the effect of roughness parameters on scattering coefficient. Oh model (1992), Dubois model (1995), modified SPA model and the reformulated Simplistic incidence angle approach (SIA) have been applied to retrieve the surface parameters. SIA gave better results than the existing models. The model suggested is the simple and efficient approach for such type ofretrieval. Chapter VI incorporates the variation of soil moisture along with the variation in surface roughness parameters for bistatic observation. It presents a critical analysis of the nature of the three soil parameters (i.e., soil moisture, horizontal roughness and the vertical roughness) on the sensitivity of scattering coefficient in the specular direction. All the four models as introduced in chapter V have been used for retrieval of soil parameters and compared. The main aim of this work is to analyze the effect of correlation length with rms height at different moisture contents for a bare soil surface. Bistatic configuration used for this purpose was the same as used for chapter V. However, the observation set was increased to 1350 for the detailed study of the microwave response at the different field parameters (i.e., correlation length (/), rms height (s) and volumetric moisture content (mv)). Nine moisture levels (0.072-0.228 cm3 cm"3) with each moisture level corresponding to three correlation lengths and five rms heights (0.40-0.88 cm) have been considered. An incidence angle of60° was found suitable for studying the individual and composite effect of soil parameters on scattering coefficient for VV polarization. It was observed that for higher and moderate moisture contents, dynamic range of specular scattering decreases and shows less variability for moderate correlation lengths. However, dynamic range remains higher for higher correlation lengths. This study is beneficial in soil moisture retrieval modeling. It provided with the detailed analysis of the mutual and individual effect of roughness parameters and soil moisture content on bistatic specular scattering coefficient. Chapter VIIpresents the summary of contributions made in the thesis and the future scope of the work.