Registration Number
0224U033045,
(0121U113718)
, R & D reports
Title
Methods, models and technologies of the data processing in the computer systems of the common and special appointment
popup.stage_title
Методи, моделі та технології обробки даних в комп’ютерних системах загального та спеціального призначення
Head
Aleksieiev Mykhailo O., д.т.н.
Registration Date
16-12-2024
Organization
National Technical University Dnipro Polytechnic
popup.description1
Research, creation and improvement of methods, models and technologies of the data processing in the computer systems of the common and special appointment
popup.description2
The research is based on the methods of the theory of mathematical modeling for the mathematical formulation of the problem and the development of methods for its solution; graph theory for the development of the architecture of the system of classification and identification of paintings; the theory of artificial neural networks for the development of an intelligent system for the classification and identification of paintings, the theory of processing experimental data and the theory of decision-making for processing model and experimental data. Experimental studies were conducted in laboratory conditions and on real objects. The effectiveness of the proposed solutions was investigated during experiments with an image database containing photos of paintings by world, European, and Ukrainian artists. Algorithm configuration parameters were found that ensure high accuracy of classification and identification
Proposes an integrated approach to processing high spatial resolution images based on the use of artificial intelligence and classical image processing algorithms. The use of the approach allows you to automate the process of semantic segmentation and analysis of multi-channel data of high spatial resolution, improve the quality of further recognition and monitoring of objects on the earth's surface. A convolutional network architecture has been developed for the tasks of semantic segmentation of satellite images of high spatial resolution, a methodology for preparing a set of input data and step-by-step implementation of cascades of network layers, an algorithm for its training have been proposed, and the choice of a training sample has been justified. Semantic features are used to reduce semantic contour level errors.
Product Description
popup.authors
Ivanchenko Oleh V.
Ishchuk Pavlo O.
Berdnyk Mykhailo H.
Holinko Oleksander V.
Hulina Iryna H.
Kabak Leonid V.
Kuvaiev Volodymyr M.
Laktionov Ivan S.
Martynenko Andrii A.
Meshcheriakov Leonid I.
Moroz Borys I.
Moroz Dmytro M.
Prykhodchenko Serhii D.
Rodna Kateryna S.
Spirintsev Viacheslav V.
Udovyk Iryna M.
Khar Alona T.
Shvachych Hennadii H.
Shevtsova Olha S.
Shyrin Artem L.
Shcherbyna Pavlo O.
popup.nrat_date
2024-12-16
