Перегляд за Автор "Salii, Yevhenii"

Зараз показуємо 1 - 6 з 6
  • Ескіз
    ДокументВідкритий доступ
    A multimodal dataset for forest damage detection and machine learning
    (Assembled by Conference Management Services, Inc., 2024) Yailymova, Hanna; Yailymov, Bohdan; Salii, Yevhenii; Kuzin, Volodymyr; Shelestov, Andrii; Kussul, Nataliia
    Accurately recognizing areas of forest damage is crucial for planning, monitoring recovery processes, and evaluating environmental impact following catastrophic events. The widespread accessibility of satellite data, coupled with the ongoing advancement of machine and deep learning techniques, as well as computer vision methods, renders the implementation of these approaches in the automatic detection of damaged forest areas highly difficult. Nevertheless, a significant challenge in this regard is the scarcity of labeled data. The purpose of this article is to provide a useful and reliable dataset for territory of Ukraine for scientists, conservationists, foresters and other stakeholders involved in monitoring forest damage and its consequences for forest ecosystems and their services. The created dataset contains 18 locations with a time series of satellite images with a resolution of up to 10 m per pixel across Ukraine, as well as weather information. The data was collected from the Copernicus Sentinel-1,2 satellite missions as well as based on ERA-5 weather information.
  • Ескіз
    ДокументВідкритий доступ
    Current Advances on Cloud-Based Distributed Computing for Forest Monitoring
    (Springer, 2023) Shelestov, Andrii; Salii, Yevhenii; Hordiiko, Nataliia; Yailymova, Hanna
    One of the most important tasks related to environmental protection is forests monitoring. Meanwhile, specialists deal with the problem of big data and the need to utilize powerful computing resources that are not always available. Cloud solutions (CREODIAS, Google Earth Engine, etc.) provide instant satellite data access and the ability to quickly and conveniently process geospatial data in the cloud and use it to search for information products. Forest monitoring is supported by the European Commission (EU project SWIFTT), the World Wildlife Fund and others. This work analyzes Sentinel-2 satellite spectral channels, which distribution of pixel values was constructed for diseased and healthy forests, and the possibility of separating these two classes was analyzed based on the Bhattacharya distance. The informativeness of time series application of the normalized difference vegetation index (NDVI) was analyzed. The assumption that the average value of NDVI decreases and the standard deviation increases when the forest changes is confirmed. Getting results for large areas will lead to a big data problem. Therefore, the structure of the pilot information system is proposed as the basis for a further cloud solution with the development of a machine (deep) learning model for forest monitoring in any territory (including Ukraine). This system allows monitoring forests dynamics based on time series of satellite data at the country level and worldwide. This will be an important step for Ukraine as a potential member of the EU in the field of providing information services and monitoring the most sensitive natural resources.
  • Ескіз
    ДокументВідкритий доступ
    Features’ Selection for Forest State Classification Using Machine Learning on Satellite Data
    (Assembled by Conference Management Services, Inc., 2024) Salii, Yevhenii; Kuzin, Volodymyr; Lavreniuk, Alla; Kussul, Nataliia; Shelestov, Andrii
    This paper discusses the use of advanced computer vision and artificial intelligence techniques for analysing remote sensing data, specifically focusing on the semantic segmentation of forest areas. The goal is to identify forest damage caused by insect pests using multispectral images from Sentinel-2 satellites. The proposed approach involves using genetic algorithms to automatically select informative features based on vegetation indices. A new fitness function is introduced to assess the quality of the selected feature sets. The neural network is then trained and tested using real data. The results of the study show the effectiveness of proposed approach and highlight its advantages over traditional methods. The developed technique allowed to obtain highly informative set of features with minimized redundancy within huge feature space with moderate amount of computation.
  • Ескіз
    ДокументВідкритий доступ
    Machine learning models and technology for classification of forest on satellite data
    (2023) Salii, Yevhenii; Kuzin, Volodymyr; Hohol, Anton; Kussul, Nataliia; Yailymova, Hanna
    The paper deals with the problem of semantic segmentation of satellite imagery to deliver forest type map with high resolution. To solve the problem, we propose 4 machine learning models. Two of them are based on Random Forest (RF) and other two - on Convolutional Neural Network (CNN) - U-Net. As an input we use 2 images of Sentinel-2 (one for summer and one for winter, 4 spectral bands from each). As an output (labels) we use the Copernicus Forest Type dataset for 2018 year. Our models showed promising results on validation data. Of all models the one based on U-Net ended up being the most efficient in forest classification with overall accuracy 91.7%. At the same time the best RF model scored with 86.5%. After comparing the results, in order to check our model transferability we created and compared forest map of northern part of Kyiv region of 2018 and 2022. The experiment confirmed the robustness of the model and it's scalability. The developed models have been implemented in the cloud platform specialized on satellite data - CREODIAS. The developed map can provide valuable data for foresters, biologists, or other researchers to make decisions about forest management and conservation, as well as to ensure that Europe's forests are managed in an ecologically sustainable way.
  • Ескіз
    ДокументВідкритий доступ
    Single-Polarized SAR Image Preprocessing in Scope of Transfer Learning for Oil Spill Detection
    (IEEE IM/CS/SMC, 2024) Kussul, Nataliia; Kuzin, Volodymyr; Salii, Yevhenii; Yailymov, Bohdan; Shelestov, Andrii
    This study proposes a novel preprocessing approach for improving oil spill detection from Synthetic Aperture Radar (SAR) satellite imagery using deep learning models. A transfer learning approach with the LinkNet segmentation architecture pre-trained on ImageNet is employed. The model is trained on Sentinel-1 SAR data from 2018–2023 using a designed preprocessing pipeline that converts the single-channel SAR input to a 3-channel RGB image. The proposed preprocessing involves transforming the original SAR intensity values to a normal distribution, extracting nonlinear features, and encoding them into the RGB channels. Quantitative results on a test set show the preprocessed model achieves an improvement of 0.038 in F1-score and 0.054 in Intersection over Union compared to the original dB-scale preprocessing approach. Qualitative evaluation on independent SAR scenes from the Mediterranean Sea also demonstrates the model's ability to generalize to new geographic areas after training on data from other regions. The proposed preprocessing technique shows promising performance gains for automatic oil spill segmentation from SAR imagery and potential for integration with other preprocessing methods and task-specific neural network architectures.
  • Ескіз
    ДокументВідкритий доступ
    Transfer learning and single-polarized SAR image preprocessing for oil spill detection
    (Elsevier, 2025) Kussul, Nataliia; Salii, Yevhenii; Kuzin, Volodymyr; Yailymov, Bohdan; Shelestov, Andrii
    This study addresses the challenge of oil spill detection using Synthetic Aperture Radar (SAR) satellite imagery, employing deep learning techniques to improve accuracy and efficiency. We investigated the effectiveness of various neural network architectures and encoders for this task, focusing on scenarios with limited training data. The research problem centered on enhancing feature extraction from single-channel SAR data to improve oil spill detection performance. Our methodology involved developing a novel preprocessing pipeline that converts single-channel SAR data into a three-channel RGB representation. The preprocessing technique normalizes SAR intensity values and encodes extracted features into RGB channels. Through an experiment, we have shown that a combination of the LinkNet with an EfficientNet-B4 is superior to pairs of other well-known architectures and encoders. Quantitative evaluation revealed a significant improvement in F1-score of 0.064 compared to traditional dB-scale preprocessing methods. Qualitative assessment on independent SAR scenes from the Mediterranean Sea demonstrated better detection capabilities, albeit with increased sensitivity to look-alike. We conclude that our proposed preprocessing technique shows promise for enhancing automatic oil spill segmentation from SAR imagery. The study contributes to advancing oil spill detection methods, with potential implications for environmental monitoring and marine ecosystem protection.