ANALYSES OF HYBRID-POLSAR DATA FOR LANDCOVER CLASSIFICATION

Abstract

The knowledge of landcover or landuse becomes increasingly important to overcome the problems related to uncontrolled urban development and depletion of valuable wetlands, agriculture lands, and forest, etc. Landcover resources are directly utilized by humans and subsequently the change in landcover directly affects the lives of human beings. For better use of the land, the information about the existing landcover patterns and the changes in landcover through time are of prime concern. Hence, landcover monitoring through the classification of landcover types has become popular among the researchers working related to earth observatory applications. To perform the landcover classification, spaceborne and airborne Synthetic Aperture Radar (SAR) systems are found to be very efficient due to its day-and-night and all-weather monitoring capabilities. Especially, the SAR system adorned with full-polarimetric (full-pol) configuration have provided promising results in past three decades. The full-pol SAR systems transmit two orthogonal polarizations in one time slot and coherently receive dual orthogonal polarizations. The information acquired by full-pol system allows to generate the full-pol scattering matrix which leads to complete-polarimetric-scattering information about the target. However, the full-pol SAR systems suffer from certain limitations. Transmission of two pulses in one time slot reduces the round-trip delay of backscattered echo by a factor of two, to avoid range ambiguities. Consequently, the swath coverage area of full-pol system is halved, which in turn has adverse impact on the revisit time. This makes full-pol systems not preferable for the applications requiring frequent Earth observation. The full-pol systems also have disadvantages related to high transmitted power requirement and increased data volume per pixel. To overcome these disadvantages of full-pol SAR systems, dual-pol SAR is being utilized. There are two types of dual-pol configurations, i.e. coherent and non-coherent, reported in literature. Between these two, the coherent type of dual-pol (compact-pol) SAR systems are preferred because it retains the relative phase difference between the two received channels to fully characterize the backscattered field. Based on the different combination of transmitted and received polarizations, three different compact-pol configurations, namely p/4, Dual-Circular-Pol (DCP), and hybrid-pol are established in literature. The p/4 mode transmits 45 oriented linear wave and coherently receives two orthogonal linear H and V polarizations, whereas, the DCP mode transmits right/left hand circular polarization and coherently receives right and left hand circular polarizations. The hybrid-pol mode is a combination of circular and linear polarizations, which transmits right/left hand circular polarization and coherently receives two orthogonal linear H and V polarizations. Further among these three compact-pol modes, the hybrid-pol is established as an optimum choice, due to having simple architecture, in-dependency on the oriented target structures, and negligible co- and cross-polarization interference. Initially, the hybrid-pol based satellites, such as Mini-SAR on India’s lunar Chandraayan-1 and Mini-RF on NASA’s Lunar Reconnaissance Orbiter were launched for the application of planetary exploration. Moreover, for Earth-observing purposes, most of the major space agencies have either launched their own hybrid-pol based satellite, or are actively indulged in establishing a new one to launch in near future. For example, Radar Imaging Satellite-1 (RISAT-1) by ISRO and Advanced Land Observation Satellite-2 (ALOS-2) by JAXA are launched in April 2012 and May 2014, respectively. Further, RADARSAT Constellation Mission (RCM) by Canadian Space Agency, SAOCOM-1 by CONAE, and NISAR jointly by NASA-ISRO, are planned to be launched in the near future. However, for the analysis of hybrid-pol data, there had been quite a few approaches reported in the open literature and further establishment of new methods is popular among the researchers. The work presented in this thesis adds on to the current research and provides four new analytical methods. These proposed methods have better performance than the conventional as well as state-of-the-art reported methods.