Vol. 9, No 2: 80–89.

Computer and Information Sciences

 2024

Scientific article

UDK 004.93
pdf-version

Maksim P. Pavlov
bachelor’s degree, Petrozavodsk State University
(Petrozavodsk, Russia),
maksim_pavlov_2003@list.ru

Semantic Segmentation of Satellite Images of the Republic of Karelia Using Spectral Analysis with a Convolutional Neural Network

Scientific adviser:
Aleksey G. Marakhtanov
Reviewer:
Lukashenko Oleg
Paper submitted on: 03/14/2024;
Accepted on: 06/29/2024;
Published online on: 06/30/2024.
 Abstract. Remote sensing images are one of the key sources of data for landscape mapping, managing natural objects, and studying environmental changes. This research explores the use of deep learning algorithms for the semantic segmentation of satellite images of the Republic of Karelia. The author investigates the hypothesis that utilizing specific combinations of light channels and spectral indices can improve the accuracy of predictions made by a neural network model. The research underscores the significance of input data selection for improving the neural network's understanding of the scene.
Keywords: Earth remote sensing, artificial intelligence, spectral image of the Earth, video analytics of satellite images, semantic segmentation of area, U-Net

For citation: Pavlov, M. P. Semantic Segmentation of Satellite Images of the Republic of Karelia Using Spectral Analysis with a Convolutional Neural Network. StudArctic forum. 2024, 9 (2): 80–89.

References

Zarubin O.A. The use of neural networks for analysis remote sensing data. Modern Scientific Researches and Innovations, 2016, No. 8. URL: https://web.snauka.ru/issues/2016/08/70887 (Accessed: 04.03.2024) (In Russ.)

Pavlov M.P. The potential of using satellite images for intelligent video analytics in the Republic of Karelia. Digital technologies for education, science, and society: proceedings of the XVII all-Russian scientific and practical conference, Petrozavodsk, 2023, pp. 77-80. (In Russ.)

Solovyev R.A., et al. Automatic segmentation of satellite imagery based on the modified UNET convolutional neural network. Engineering Journal of Don, 2017, No. 4. URL: http://www.ivdon.ru/ru/magazine/archive/n4y2017/4433 (Accessed: 04.03.2024) (In Russ.)

Tabunshchyk V.A., Petlukova E.A., Hytrin M.O. The use of satellite imagery Sentinel-2 for analysis of land used in agriculture (for example Razdolnensky District of the Republic of Crimea). Proceedings of the T.I. Vyazemsky Karadag Scientific Station – Nature Reserve of the Russian Academy of Sciences, 2018, No. 1, pp. 43-57. (In Russ.)

Khryashchev V.V., Priorov A.L., Pavlov V.A., et al. Modification of the U-Net convolutional neural architecture in the multi-channel satellite images segmentation problem. Digital Signal Processing, 2019, No. 3, pp. 39-45. (In Russ.)

Cristea C., Jocea A.F. Use of Sentinel-2 data for forest inventory. 18th International Multidisciplinary Scientific GeoConference SGEM 2018, 2018, pp. 713-718. DOI: 10.5593/sgem2018/2.3/S11.090.

Peuch V.H., Engelen R., Ades M., et al. The use of satellite data in the Copernicus atmosphere monitoring service (CAMS). International Geoscience and Remote Sensing Symposium (IGARSS), 2018, pp. 1594-1596. DOI: 10.1109/IGARSS.2018.8518698.

Semeraro T., Luvisi A., Mastroleo G., et al. Modelling fuzzy combination of remote sensing vegetation index for durum wheat crop analysis. Computers and Electronics in Agriculture, 2019, Vol. 156, pp. 684-692. DOI: 10.1016/j.compag.2018.12.027.

Singh N. J., Nongmeikapam K. Semantic segmentation of satellite images using deep-UNet. Arabian Journal for Science and Engineering, 2023, Vol. 48, pp. 1193-1205. DOI: 10.1007/s13369-02-06734-4.

Displays: 135; Downloads: 32;