Search academic texts

Information × Registration Number 0214U000223, 0112U000546 , R & D reports Title Optimal method of recognition and classification of objects and bundled software of automatic recognition of space information of radio popup.stage_title Head Dymin Olexandr, Registration Date 24-01-2014 Organization Kharkov National University named after V.N. Karazin popup.description2 Object of investigation: cover of the Earth's surface, Earth's natural resources and objects that can cause man-made and natural disasters that can be studied by the methods of remote sensing and space shooting of an environment on the data of multispectral electromagnetic radiation of the surface of objects in the range of the spectrum from radio waves to ultraviolet. The aim of the work: to provide quality improvement of object recognition of remote sensing. The method of investigations: standard methods of mathematical physics and the original author methods of solving ill-posed problems, methods of approximation of functions to compress the input data, recognition methods on the probabilistic learning sample, methods of probability theory and fuzzy sets, EM algorithm, methods of Fisher information matrix. Based on the probability of learning samples, the EM algorithm, the method of maximum likelihood and Fisher information the optimal method of recognition and classification of objects on data of space observations obtained by remote sensing at different wavelengths was developed, where the criteria of classification quality is maximized in two main factors: the characteristics of objects and rules for the classification of a family of mixed probability distribution models of recognition characteristics of objects. The computer software system RIKDEDIN was developed for creating the artificial simulation data of multichannel remote sensing, multi-dimensional data visualization for interactive statistical analysis, classification of the objects of learning set, the calculation of the Fisher information, the distinctive characteristics of objects, and dynamic model of recognition. The results may be of interest to organizations working in the space industry and dealing with remote sensing, and are intended to improve the quality of processing and classification of the acquired data. RECOGNITION AND CLASSIFICATION, PROBABILISTIC LEARNING SAMPLE, FISHER INFORMATION, REMOTE SENSING Product Description popup.authors Ахмедов Р. Думіна О. Катрич В. Колтунов Й. Ляховський А. Стаднік Ю. Хмара С. popup.nrat_date 2020-04-02 Close