Multidimensional SAR satellite images - a mapping perspective

Abstract

In this thesis, the use of Synthetic Aperture Radar (SAR) satellite images in potential mapping applications areas in Finland was studied. SAR is an active sensor using the microwave region of the electromagnetic spectrum in its pulses. Microwaves penetrate clouds, smoke and dust without noticeable attenuation, enabling all-weather and night-and-day satellite imaging. Because cloudiness is very common in Finland, SAR is of importance for monitoring purposes. Recently, the number of SAR satellites has increased notably. First, SAR images can now be acquired more frequently than before. Second, SAR images can be acquired in multiple polarization channels, different frequency bands and various imaging geometries using several satellites. As the dimensionality of SAR data increases, it can be expected that more automatic and sophisticated processes are needed. The objective was to study streamlining of mapping processes based on SAR satellite images. First, automatic matching of remote sensing images and existing vector maps was studied. Second, multidimensional SAR satellite images were used in selected mapping applications. Example cases include agricultural monitoring (dual-polarimetric Envisat ASAR), detection of buildings (Radarsat-1 SAR), detection and verification of building subsidence (ERS-1 and ERS-2 SAR), and forest biomass mapping (ALOS PALSAR). The results showed that it may be possible to use existing vector maps to refine the geocoding parameters of SAR images. According to the ground check points, accuracy of around 2 pixels was achieved in the image-to-map co-registration. When Persistent Scatterers SAR Interferometry (PSI) subsidence rates of individual buildings were compared with the levelling measurements, RMSE of 0.82 mm/year was achieved. At its best R² values of 0.55 and 0.72 were obtained for crop biomass and forest above ground volume estimations respectively. Crop species were classified with the overall accuracy of 75%. Building detection percentages varied between 13% and 98%, depending on the orientation of the building wall and building height. The author believes that the results might show commercial potential and have a socioeconomic impact, providing the prices of SAR images decrease. PSI could be used to monitor subsidence of urban areas operationally. SAR satellite is the only way to monitor wide agricultural areas in Finland, even though the results are somewhat poor. In the case of forests, SAR enables more frequent update of forest information when compared to airborne laser scanning. SAR images might have potential in detection of changes in urban areas, especially in the remote areas of the world. The future of SAR satellites appears promising because more satellite systems will be launched in the next ten years.