Kavats Y. Information technology for processing and interpretation optical and radar satellite images

Thesis for scientific degree of Candidate of Technical Sciences in specialty 05.13.06— «Information technology». (05 – Technical Sciences). – National Metallurgical Academy of Ukraine, National Technical University «Dnipro Polytechnic», Dnipro, 2020. The important scientific and applied problem of increasing the efficiency of methods for interpretation and operational monitoring of objects on the earth's surface by developing new information technologies for interpretation and multidimensional processing of optical and radar satellite images is solved in the thesis. Information technologies of satellite data of different physical nature (optical and radar) preprocessing and thematic processing have been developed. The interpolation algorithm of multispectral aerospace images is improved based on wavelet transforms, which in comparison with the classical interpolation algorithms allows to achieve better quality of results. The scalable image is more informative. Studies have shown that the value of information entropy is higher than corresponding values obtained using known interpolation algorithms. Information technology for enhancing the spatial resolution of multispectral aerospace images based on linear IHS-transformation is improved and its influence is determined. It has been investigated that as a result of multispectral images processing using linear and nonlinear IHS-transformations, the correlation is reduced, the errors are reduced, the synthesized images have increased informative content compared to the primary images. The information and signal entropy of images synthesized using IHS-conversion significantly exceeds the corresponding values calculated for the primary multispectral image. Visual and quantitative analysis convincingly demonstrates that, unlike nonlinear conversion, image synthesized by linear IHS-model results in color distortions and unrealistic perception of multispectral image. The basic steps of automated method of preprocessing Sentinel-1 radar satellite data for solving applied monitoring problems are analyzed. Information technology has been developed for shadow detection and compensation on geospatial multispectral satellite images of ultra-high spatial resolution. The obtained results show that the proposed technology for shadow detection and compensation on ultra-high spatial resolution multispectral satellite images allows to qualitatively compensate shadow areas unlike other algorithms and to avoid the formation of erroneous objects. Information technology of artificial objects interpretation on high spatial resolution multispectral photogrammetric images based on morphological indices is proposed. A comparative analysis of efficiency applying different types of structural element for buildings recognition on high spatial resolution multispectral aerospace images based on morphological indices is performed. The research results indicate that the use of the LINE structural element allows to achieve the best results compared to other types, effectively detect boundaries and contours of buildings with complex shape. A comparative analysis of the efficiency of classical methods of photogrammetric images classification is performed. Studies have shown that the use of information technology to artificial objects interpretation based on morphological indices allows to clearly find building roofs in comparison with classic classification algorithms. The accuracy is 90%. The developed information technologies for solving practical problems of optical and radar satellite monitoring allow detecting changes and temperatures of anthropogenic objects based on different-time images, interpret changes in forest plantations and floods. The proposed information technologies are based on temporal changes, allowing observation and research in large-scale inaccessible territories over a long period. The automation of the proposed methods and information technologies for solving satellite monitoring applications is implemented in the form of a cloud-based «EOS Processing» web platform that allows to perform operations on detecting changes in the earth's surface using optical and radar data without uploading data to local computing capacities. The speed of processing operations run with the help of web-service «EOS Processing» in comparison with the full cycle of stationary processing is high due to automation and cloud computing using distributed processing of information. Keywords: space data, satellite images, information technology, interpretation, recognition, web-service, monitoring, time series data.