Unsupervised image segmentation methods for mapping stress-affected regions of agricultural land
DOI:
https://doi.org/10.24867/16BE03PajevicKeywords:
Segmentation, machine learning, computer vision, Sentinel-2Abstract
The paper describes the segmentation of Sentinel-2 satellite images based on derived vegetation indices in order to map stress-affected regions of agricultural land. The database consists of a time series of multispectral images of 48 agricultural parcels in Vojvodina. The time series includes images for 13 dates in the period from March to September 2020. Three segmentation methods were applied: threshold segmentation, segmentation using the k-means algorithm, and segmentation using the pyImSegm library. The first damaged area was not successfully detected by segmentation with the first method, while the other two methods showed a good ability to map less developed regions.
References
[1] Zheng, Q., Huang, W., Cui, X., Shi, Y. and Liu, L., 2018. New Spectral Index for Detecting Wheat Yellow Rust Using Sentinel-2 Multispectral Imagery. Sensors, 18(3), p.868.
[2] Sentinel-2 User Handbook, https://sentinel.esa.int/documents/247904/685211/Sentinel-2_User_Handbook
(pristupljeno u septembru 2021.)
[3] Chemura, A., Mutanga, O. and Dube, T., 2016. Separability of coffee leaf rust infection levels with machine learning methods at Sentinel-2 MSI spectral resolutions. Precision Agriculture, 18(5), pp.859-881.
[4]https://stanford.edu/~cpiech/cs221/handouts/kmeans.html (pristupljeno u septembru 2021.)
[5]https://www.mathworks.com/help/images/ref/superpixels.html (pristupljeno u septembru 2021.)
[6] Borovec, J., Švihlík, J., Kybic, J. and Habart, D., 2017. Supervised and unsupervised segmentation using superpixels, model estimation, and graph cut. Journal of Electronic Imaging, 26(06), p.1.
[2] Sentinel-2 User Handbook, https://sentinel.esa.int/documents/247904/685211/Sentinel-2_User_Handbook
(pristupljeno u septembru 2021.)
[3] Chemura, A., Mutanga, O. and Dube, T., 2016. Separability of coffee leaf rust infection levels with machine learning methods at Sentinel-2 MSI spectral resolutions. Precision Agriculture, 18(5), pp.859-881.
[4]https://stanford.edu/~cpiech/cs221/handouts/kmeans.html (pristupljeno u septembru 2021.)
[5]https://www.mathworks.com/help/images/ref/superpixels.html (pristupljeno u septembru 2021.)
[6] Borovec, J., Švihlík, J., Kybic, J. and Habart, D., 2017. Supervised and unsupervised segmentation using superpixels, model estimation, and graph cut. Journal of Electronic Imaging, 26(06), p.1.
Downloads
Published
2022-01-25
Issue
Section
Electrotechnical and Computer Engineering
