KFU. Card publication. Recognition of Slightly Blurred Images of Planets by a Convolutional Neural Network

KAZAN
FEDERAL UNIVERSITY

- RUS
- 中文
- ES

RECOGNITION OF SLIGHTLY BLURRED IMAGES OF PLANETS BY A CONVOLUTIONAL NEURAL NETWORK

Form of presentation

Conference proceedings in international journals and collections

Year of publication

2024

��

��

Authors, employees of KFU

Egorchev Anton Aleksandrovich, author

Tumakov Dmitriy Nikolaevich, author

Bibliographic description in the original language

Tuliabaeva, D. Recognition of Slightly Blurred Images of Planets by a Convolutional Neural Network / D. Tuliabaeva, D. Tumakov, A. Egorchev // III International Scientific and Practical Conference Technologies, Materials Science and Engineering (EEA-III-2024) : AIP Conference Proceedings, Dushanbe. Vol. 3243. – Melville: AIP PUBLISHING, 2024. – P. 20087. – DOI 10.1063/5.0247350

Annotation

AIP Conference Proceedings. �� 3243. Melville, 2024

Keywords

Image classification, Convolutional Neural Network

The name of the journal

AIP Conference Proceedings. �� 3243. Melville, 2024

Please use this ID to quote from or refer to the card

https://repository.kpfu.ru/eng/?p_id=317401&p_lang=2

Full metadata record

Field DC	Value	Language
dc.contributor.author	Egorchev Anton Aleksandrovich	ru_RU
dc.contributor.author	Tumakov Dmitriy Nikolaevich	ru_RU
dc.date.accessioned	2024-01-01T00:00:00Z	ru_RU
dc.date.available	2024-01-01T00:00:00Z	ru_RU
dc.date.issued	2024	ru_RU
dc.identifier.citation	Tuliabaeva, D. Recognition of Slightly Blurred Images of Planets by a Convolutional Neural Network / D. Tuliabaeva, D. Tumakov, A. Egorchev // III International Scientific and Practical Conference Technologies, Materials Science and Engineering (EEA-III-2024) : AIP Conference Proceedings, Dushanbe. Vol. 3243. – Melville: AIP PUBLISHING, 2024. – P. 20087. – DOI 10.1063/5.0247350	ru_RU
dc.identifier.uri	https://repository.kpfu.ru/eng/?p_id=317401&p_lang=2	ru_RU
dc.description.abstract	AIP Conference Proceedings. �� 3243. Melville, 2024	ru_RU
dc.description.abstract	The influence of the degree of blur on accuracy of recognition of celestial bodies images is studied. For this purpose, color photographs of nine celestial objects of the Solar System are used including eight confirmed planets, two dwarf planets and the Earth's satellite - the Moon. The training and test datasets are represented by images of celestial bodies against a black background measuring 256 by 144 pixels. The analysis is carried out on color and corresponding black-and-white images. It is found that by increasing the filter size and adding additional convolutional layers to the neural network, recognition accuracy increases. The study confirms that the accuracy of celestial body recognition is highly dependent on even the degree of blur (3x3 matrix), decreasing from 99% to 79% for color images and from 97.5% to 40% for grayscale images. It is shown that a dataset enlarged by adding images obtained through rotations by 45, 90, 180 and 270 degrees leads to a significant (by about 20%) increase in the accuracy of the model for blurred images. Histograms are presented demonstrating recognition accuracy depending on the degree of blur.	ru_RU
dc.language.iso	ru	ru_RU
dc.subject	Image classification	ru_RU
dc.subject	Convolutional Neural Network	ru_RU
dc.title	Recognition of Slightly Blurred Images of Planets by a Convolutional Neural Network	ru_RU
dc.type	Conference proceedings in international journals and collections	ru_RU