Lysenko A. Synthetic-aperture multi-polarization radar data informativity enhancement technique

In modern world of micro-, macroeconomy the ability for risks, status, results control, assessment or forecast is necessary. To supply growing demands big amounts of relevant data must be provided. Owing to recent satellite coverage and remote sensing methods, it is possible to retrieve huge extent of land objects data, practically, from any point of the land surface in a remarkably short time. An important qualitative characteristic of Earth remote sensing data is their informativeness, which is responsible for the amount of useful information for solving a specific thematic problem. Therefore, this parameter is different for different tasks. Taking into account growing needs for high quality data in order to obtain more accurate results, satellite imagery informativity enhancement is a relevant task of nowadays. Thus, the aim of this work is to increase the informativity of space images obtained with a multi-polarization synthetic-aperture radar (SAR), by using modern exploratory works to develop the appropriate technique. To achieve the given goal, it is envisaged to perform several scientific tasks, the first of which was the analysis of existing methods for satellite imagery informativity enhancement, and the relationship between informativity and spatial resolution establishment. The satellite images obtained from the constellation of satellites of the Sentinel-1 radar system were chosen as the source SAR data. Since this satellite provides data in two polarizations at once, one of which coincides in the emitted and received polarization modes of the electromagnetic pulse, and the other contains an interpolarization component, the models of satellite radar backscattering signal were developed to transform different polarization data into a common physical quantity – dielectric permittivity of the land surface. In order to enhance the spatial resolution, a corresponding mathematical model and algorithm for the low spatial resolution satellite images fusion into a single two-fold enlarged image of enhanced spatial resolution, were developed. The assessment of the achieved result – radar data informativity enhancement – was carried out quantitatively by comparing the actual spatial resolution of the input and output images. In turn, the actual spatial resolutions were estimated by the methods of spatial-frequency analysis through the approximate modulation transfer functions (MTF) of digital images. Scientific novelty For the first time a technique for multi-polarization synthetic-aperture radar data informativity enhancement technique was developed. The technique provides an enhancement in the spatial resolution of the output radar image due to the joint subpixel processing of several input images obtained in each polarization separately and converted into a single physical value inherent in the land surface – dielectric permittivity, roughness, moisture content, and so on. As a result of such transformation, it becomes possible to correctly apply the superresolution algorithm to a set of different polarization radar images. The model for converting multi-polarization radar data into dielectric permittivity of the land surface has been improved by imposing physically determined limitations of the target value and applying iterative forward-inverse modeling of the reflected radar signal. The algorithm for restoring a joint image of subpixel resolution from a set of satellite images has been improved by its adaptation to the radar data, which consists in applying additional speckle filtering directly in the frequency domain, where all other processing is performed. Practical implications According to the results of experimental testing on many actual dual-polarization radar images, the developed technique provided, on average, an enhancement in spatial resolution by 85.4 %. In addition to actually enhance the informativity of multi-polarization radar imaging materials, the proposed technique provides restoration of the spatial distribution of the target physical characteristics of the land surface with enhanced resolution, which corresponds to the modern concept of ARD (analysis-ready data), which means obtaining from raw satellite data maps of physical/biophysical parameters of the land surface that are understandable to experts in applied ground-based studies. The developed technique can be used for such Earth observation applications as: environmental monitoring, geophysical mapping, mineral exploration, land degradation forecasting, climate change research, etc. Keywords: Earth remote sensing, satellite data, synthetic-aperture radar (SAR), radar backscattering coefficient, dielectric permittivity, superresolution, subpixel processing