Search academic texts

Information × Registration Number 0214U002538, 0113U003698 , R & D reports Title Development of geoinformation agriculture portal and thematic agro-services based on satellite data popup.stage_title Head Skakun Sergii, Registration Date 15-01-2014 Organization Space Research Institute popup.description2 Objective of the project is to develop and implement geoinformation technologies for agricultural monitoring using satellite and ground observations and to adapt European methods in local government. In this study, state-of-the-art classification methods (including kernel and neural network techniques) applied to a new area, in particular classification of agricultural crops using satellite data. New approaches to tackle the problem of non-representative of training set are proposed. These are based on the hypothesis according to which the conditional distributions of the signal is the same for the test and training set, and only change of a priori probabilities of classes is assumed. For the classification of crops optical and radar (multipolarization synthetic aperture radar) satellite data are applied. For this, the methods of heterogeneous satellite data fusion are developed. For radar data optimum polarizations for crop classification are identified. To collect training data we used ground survey "along the road" as well as data from the JECAM twst site in Ukraine (http://jecam.org.ua). based observations and 2 times decrease error estimates for land areas only statistics. As a result of the project in 2013 we developed and improved kernel and neural network methods for crop classification using optical and radar satellite data. The analysis of the effectiveness of the methods of classification and selection of the optimal classifier is presented. A thematic map of crops based on satellite and ground data is produced. Efficiency in solving agriculture monitoring problem is increased 2-3 times using satellite data. In particular, the use of radar satellite multipolarization data allowed us to improve the accuracy of identification of agricultural crops in comparison with optical satellite data by 10-15%, the use of satellite data in the problem of evaluating agricultural crops allowed 2 times to reduce costs required for ground-based observations and 2 times decrease error estimates for land areas only statistics. Product Description popup.authors Колотій А.В. Кравченко О.М. Кригін В.М. Остапенко В.А. Скакун С. В. Яйлимов Б.Я. popup.nrat_date 2020-04-02 Close