Search academic texts

Information × Registration Number 0218U000412, 0117U007234 , R & D reports Title Deep Leaming models and methods for geospatial analysis in agriculture domain popup.stage_title Head Andrii Shelestov, Registration Date 29-01-2018 Organization Faculty of аircraft and space systems NTUU "KPI" popup.description2 The object of research is high resolution satellite images and methods of their processing. The aim of the work is to explore and improve the models and methods of machine learning and deep learning for geospatial analysis in the agriculture domain. Method of research - machine learning, geospatial analysis, object-oriented design, structural functional analysis, statistical methods, neural networks. With launch of new modern satellite missions the development of methods of machine learning applicable for geospatial data, in particular satellite data, is becoming more and more urgent. Within this project the applicability of main machine learning methods for land cover mapping as well as use of them with high spatial resolution satellite imagery was analyzed. Particular attention is paid to deep architectures, in particular, convolutional neural networks that are the most powerful and precise method for visual images recognition at the moment. The main advantages of deep learning methods over the traditional approaches to the classification tasks that have been used over the past decades and based on expert knowledge necessary to distinguish the features from the input data have been determined within the project. The main principles of the formation of information products according to international programs and standards were investigated and a collection of in-situ and satellite data for the territory of Ukraine was performed. Product Description popup.authors Гільчук А.В. Колотій А.В. Куссуль Н.М. Лавренюк А.М. Лавренюк М.С. Мирошникова І.Ю. Циганкова О.В. Шелестов А. Ю. Яйлимов Б.Я. popup.nrat_date 2020-04-02 Close