Кафедра математичного моделювання та аналізу даних (ММАД)

Постійне посилання на фонд

https://ela.kpi.ua/handle/123456789/43446

Сайт кафедри: http://mmda.ipt.kpi.ua/

Переглянути

Зараз показуємо 1 - 20 з 226

Відкритий доступ
3D Scene Reconstruction with Neural Radiance Fields (NeRF) Considering Dynamic Illumination Conditions
(Anhalt University of Applied Sciences, 2023) Kolodiazhna, Olena; Savin, Volodymyr; Uss, Mykhailo; Kussul, Nataliia
This paper addresses the problem of novel view synthesis using Neural Radiance Fields (NeRF) for scenes with dynamic illumination. NeRF training utilizes photometric consistency loss that is pixel-wise consistency between a set of scene images and intensity values rendered by NeRF. For reflective surfaces, image intensity depends on viewing angle and this effect is taken into account by using ray direction as NeRF input. For scenes with dynamic illumination, image intensity depends not only on position and viewing direction but also on time. We show that this factor affects NeRF training with standard photometric loss function effectively decreasing quality of both image and depth rendering. To cope with this problem, we propose to add time as additional NeRF input. Experiments on ScanNet dataset demonstrate that NeRF with modified input outperforms original model version and renders more consistent 3D structures. Results of this study could be used to improve quality of training data augmentation for depth prediction models (e.g. depth-from-stereo models) for scenes with non-static illumination.
Відкритий доступ
3D-реконструкція приміщень за сферичними панорамами
(КПІ ім. Ігоря Сікорського, 2023) Крохальов, Іван Данилович; Орєхов, Олександр Арсенiйович
Магістерська дисертація: 58 с., 35 рис., 2 табл., 15 джерел. Об`єкт дослідження – панорамні знімки приміщень, 3D-моделі приміщень, отримані внаслідок реконструкції; усі матеріали, що є результатом розв’язання проміжних задач реконструкції; Предмет дослідження – методи 3D-реконструкції, підхід Structure-from-Motion, його поведінка при різних прикладах вхідних даних, можливі кроки для покращення якості реконструкції; Мета роботи — побудувати алгоритм 3D-реконструкції на основі підходу Structure-from-Motion, для використання на різних прикладах вхідних даних. В роботі наведено програмну реалізацію математичної моделі, що дозволяє проводити дослідження у відповідності до мети роботи. Проведений порівняльний аналіз різних конфігурацій 3D-реконструкції. В роботі також наведено порівняння запропонованого метода із альтернативними підходами до 3D-реконструкції.
Відкритий доступ
A generalized model for mapping sunflower areas using Sentinel-1 SAR data
(Elsevier, 2024) Qadir, Abdul; Skakun, Sergii; Kussul, Nataliia; Shelestov, Andrii; Becker-Reshef, Inbal
Existing crop mapping models, rely heavily on reference (calibration) data obtained from remote sensing observations. However, the transferability of such models in space and time, without the need for additional extensive datasets remains a significant challenge. There is still a large gap in developing generalized classification models capable of mapping specific or multiple crops with minimal calibration data. In this study, we present a generalized automatic approach for sunflower mapping at 20-m spatial resolution, using the C-band Sentinel-1 (S1) synthetic aperture radar (SAR) data driven by previously developed phenological metrics. These metrics characterize the directional behavior of the sunflower head, capturing distinct backscattering responses in SAR data acquired from ascending and descending orbits. Specifically, we utilize SAR-derived backscatter values in VH and VV polarization, as well as their ratio VH/VV, as input features to a random forest classifier that was calibrated for the year 2022 in Ukraine. This model is further directly applied to selected sites for multiple years in Ukraine (generalization in time) and other major sunflower producing countries (generalization in space): Ukraine for 2018–2020, and Hungary, France, Russia and USA for 2018. Our results reveal that the model based on features acquired from descending orbits outperforms its ascending orbit counterpart because of the directional behavior of sunflower: user's accuracy (UA) of 96%, producer's accuracy (PA) of 97% and F-score of 97% (descending) compared to UA of 90%, PA of 89% and F-score of 90% (ascending). When generalized to other years and countries, our model achieves an F-score exceeding 77% for all cases, with the highest F-scores (>91%) observed in Ukraine and Russia sites and the lowest (77%) for the US site. We further utilize the produced maps (pixel-based) for the selected regions and years to estimate sunflower planted areas using a statistical sampling-based approach. Our estimates yield the relative root mean square error (RMSE) as 19.7% of the mean area, when compared to the reference data from official statistics and reference maps. These findings demonstrate the robustness of our proposed approach across space and time in generating accurate sunflower maps, its ability to mitigate cloud cover issues through spaceborne SAR data acquisitions, and its potential for obtaining estimates of sunflower planted areas. This research emphasizes the importance of developing interpretable and domain-specific machine learning models that can be readily extended to multiple geographical regions with little to no labelled datasets.
Відкритий 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.
Відкритий доступ
Acceleration of computations in modelling of processes in complex objects and systems
(Інститут загальної енергетики НАН України, 2024) Khaidurov, Vladyslav; Tatenko, Vadym; Lytovchenko,; Tsiupii, Tamara; Zhovnovach, Tetiana
The development of methods of parallelization of computing processes, which involve the decomposition of the computational domain, is an urgent task in the modeling of complex objects and systems. Complex objects and systems can contain a large number of elements and interactions. Decomposition allows you to break down a system into simpler subsystems, which simplifies the analysis and management of complexity. By dividing the calculation area of the part, it is possible to perform parallel calculations, which increases the efficiency of calculations and reduces simulation time. Domain decomposition makes it easy to scale the model to work with larger or more detailed systems. With the right choice of decomposition methods, the accuracy of the simulation can be improved, since different parts of the system may have different levels of detail and require appropriate methods of additional analysis. Decomposition allows the simulation to be distributed between different participants or devices, which is relevant for distributed systems or collaborative work on a project. In this work, mathematical models are built, which consist in the construction of iterative procedures for "stitching" several areas into a single whole. The models provide for different complexity of calculation domains, which makes it possible to perform different decomposition approaches, in particular, both overlapping and non-overlapping domain decomposition. The obtained mathematical models of subject domain decomposition can be applied to objects and systems that have different geometric complexity. Domain decomposition models that do not use overlap contain different iterative methods of "stitching" on a common boundary depending on the types of boundary conditions (a condition of the first kind is a Dirichlet condition, or a condition of the second year is a Neumann condition), and domain decomposition models with an overlap of two or more areas consist of the minimization problem for constructing the iterative condition of "stitching" areas. It should be noted that the obtained models will work effectively on all applied tasks that describe the dynamic behavior of objects and their systems, but the high degree of efficiency of one model may be lower than the corresponding the degree of effectiveness of another model, since each task is individual.
Відкритий доступ
Advanced Method of Land Cover Classification Based on High Spatial Resolution Data and Convolutional Neural Network
(Anhalt University of Applied Sciences, 2022) Shelestov, Andrii; Yailymov, Bohdan; Yailymova, Hanna; Shumilo, Leonid; Lavreniuk, Mykola; Lavreniuk, Alla; Sylantyev, Sergiy; Kussul, Nataliia
Відкритий доступ
Agriculture land appraisal with use of remote sensing and infrastructure data
(IEEE, 2022) Kussul, Nataliia; Shelestov, Andrii; Yailymova, Hanna; Shumilo, Leonid; Drozd, Sophia
Відкритий доступ
Air Quality as Proxy for Assesment of Economic Activity
(2023) Yailymova, Hanna; Kolotii, Andrii; Kussul, Nataliia; Shelestov, Andrii
In Ukraine most of citizens and economic activity are concentrated over urban city centers and city functional areas. Thus, Air Quality and, in particular, levels of fine particulate matter (e.g., PM2.5 and PM 10 ) over cities can be a proxy for assessment of economic activity and density of city populations. Since the russia invasion to Ukraine started on 24 of February 2022 according to UNHCR (the UN Refugee Agency) 8 million refugees from Ukraine have now been registered across the Europe. Almost 7 million more are displaced within the country. On the other hand, there is no official statistics from national statistical service showing current influence of invasion on city economic activity or inhabitants amount. Thus, such a proxy can be used to see current situation by analyzing of particulate matter time series. In this work we compare averaged annual cumulated PM2.5 for 2018–2021 years with values for 2022 and estimate the correlation them with publicly available statistics on migration to see some relations. Global Sustainable Development Goal (SDG) indicator 11.6.2, “Annual mean levels of fine particulate in cities (population weighted)” is being extended for 2022 and compared with previous years.
Відкритий доступ
Air Quality Estimation in Ukraine Using SDG 11.6.2 Indicator Assessment
(Remote Sensing, 2021-11) Shelestov, Andrii; Yailymova, Hanna; Yailymov, Bohdan; Kussul, Nataliia
Відкритий доступ
Artificial Intelligence models in solving Ill-posed Inverse problems of Remote Sensing GHG emission
(Leaving Planet Symposium, 2022) Sylantyev, Sergiy; Yailymova, Hanna; Shelestov, Andrii
Відкритий доступ
Assessing damage to agricultural fields from military actions in Ukraine: An integrated approach using statistical indicators and machine learning
(2023) Kussul, Nataliia; Drozd, Sofiia; Yailymova, Hanna; Shelestov, Andrii; Lemoine, Guido; Deininger, Klaus
The ongoing full-scale Russian invasion of Ukraine has led to widespread damage of agricultural lands, jeopardizing global food security. Timely detection of impacted fields enables quantification of production losses, guiding recovery policies and monitoring military actions. This study presents a robust methodology to automatically identify agricultural areas damaged by wartime ground activities using free Sentinel-2 satellite data. The 10 m resolution spectral bands and vegetation indices are leveraged, alongside their statistical metrics over time, as inputs to a Random Forest classifier. The algorithm efficiently pinpoints damaged fields, with accuracy metrics around 0.85. Subsequent anomaly detection delineates damages within the fields by combining spectral bands and indices. Applying the methodology over 22 biweekly periods in 2022, approximately 500 thousand ha of cropland across 10 regions of Ukraine were classified as damaged, with the most significant impacts occurring from March to September. The algorithm provides updated damage information despite cloud cover and vegetation shifts. The approach demonstrates the efficacy of automated satellite monitoring to assess agricultural impacts of military actions, supporting recovery analysis and documentation of war crimes.
Відкритий доступ
Assessing Ukraine`s solar power potential: a comprehensive analysis using satellite data and fuzzy logic
(Assembled by Conference Management Services, Inc., 2024) Drozd, Sofiia; Kussul, Nataliia
This study evaluates the land suitability for the placement of solar power stations in Ukraine, utilizing satellite data on climate factors (Global Horizontal Irradiance, temperature, precipitation, wind speed), topography (elevation, slope), and land use. Fuzzy logic, pairwise comparisons, and weighted linear combination methods were utilized to develop a high-resolution (100 m) land suitability map for the installation of solar power plants. The results show that more than half (54.5%) of Ukraine’s territory has a high suitability score (exceeding 0.65) for solar power stations, particularly in the southern and eastern regions, such as Odessa, Kherson, Mykolaiv, Zaporizhia, Donetsk, and Crimea. Only 10.68% of the land has a suitability score less than 0.6, and 18.18% is deemed absolutely unsuitable (with a score of 0, due to land cover), primarily located in the western and northern parts of the country. This indicates that Ukraine has significant potential for green energy production. The study provides an effective and useful tool for decision-making on the optimal location of solar power facilities in Ukraine.
Відкритий доступ
Assessing Ukrainian Territory Suitability for Solar Power Station Placement Using Satellite Data on Climate and Topography
(IEEE, 2023) Kussul, Nataliia; Drozd, Sofiia
This research aims to assess the suitability of Ukrainian territories for the placement of solar power stations using satellite data on climate and topographic characteristics. The suitability of the territories was determined using a weighted sum method, incorporating input parameters from climate maps sourced from ERA5- Land dataset, which included data on annual global horizontal solar irradiation (GHI), accumulated annual temperature above 25°C, average annual wind speed, and maps of accumulated annual precipitation. Additionally, topographic maps from the SRTM dataset were utilized, providing information on elevations, slopes, and terrain shading. Furthermore, data from Wikimapia on the locations of existing major solar power stations in Ukraine were used to verify the placement optimization. The results of the study revealed that the largest portion of the country (over 48%) exhibits moderate suitability scores (0.3-0.4). Favorable territories (suitability score above 0.3) outweigh unsuitable ones for solar power stations. The southern regions and the Crimean Peninsula offer the most favorable conditions for the placement of solar farms. Overall, all analyzed major solar power stations in Ukraine were located in optimal territories. Furthermore, it was found that certain regions such as Odessa, Poltava, Kharkiv, Zaporizhia, Dnipropetrovsk, Donetsk, and Luhansk demonstrate good suitability scores (0.3-0.4), yet they are not fully exploited. These regions hold significant potential for the future construction of powerful and productive solar power stations.
Відкритий доступ
Automated detection and assessment of war-induced damage to agricultural fields using satellite imagery
(Одеський національний технологічний університет, 2024) Kussul, N.; Drozd, S.; Yailymova, H.
This paper introduces a methodology based on machine learning and remote sensing for detecting military-induced damages to agricultural lands in Ukraine using free Sentinel-2 satellite data. The most informative spectral bands (B2, B3) and vegetation indices (NDVI, GCI) were experimentallyselected for recognizing damaged fields through the Random Forest classification algorithm. Additionally, an anomaly detection method based on the estimation of deviations of pixel values from the mean within each field was applied to determine local damage in the identified affected fields. The proposed methodology demonstrated high classification accuracy with an f1-score of 0.87%, producer’s accuracy of 0.89%, user’s accuracy of 0.85, and sensitivity for detecting local damage. The developed anomaly detection method allows to recognize damage visible on the 10-meter pixel of the Sentinel-2 satellite, but does not identify small craters.Cloudiness of satellite images can significantly impair the accuracy of damage detection, and the method of local damage detection can respondto non-military anomaliesand requires careful selection of threshold coefficients for each field. The study conducted a comprehensive assessment of damages inflicted on Ukrainian agricultural fields during the period 2022-2023, revealing that a total of 1,544,952 hectares, equivalent to 5.72% of the total agricultural area, experienced damage. This included 509,107 ha of wheat, 114,302 ha of sunflower, 68,830 ha of maize, 4,029 ha of rapeseed, and 16,561 ha of other crops. The most affected regions were Donetsk, Zaporizhia, and Kherson oblasts. The comprehensive findings of this research provide valuable insights for monitoring the state of agriculture and formulating strategic plans for the recovery of agricultural resources amidst the ongoing military conflict.
Відкритий доступ
Automatic Deforestation Detection based on the Deep Learning in Ukraine
(IEEE, 2021) Shumilo, Leonid; Lavreniuk, Mykola; Kussul, Nataliia; Shevchuk, Bella
Відкритий доступ
Autoregressive models for air quality investigation
(2023-08) Zalieska, Olena; Yailymova, Hanna
The aim of the work is to build a forecast of air quality in Kyiv for some period of time. For this purpose we preprocessed and analized data, selected and fitted a model.
Відкритий доступ
Biophysical Impact of Sunflower Crop Rotation on Agricultural Fields
(Sustainability, 2022) Kussul, Nataliia; Deininger, Klaus; Shumilo, Leonid; Lavreniuk, Mykola; Ayalew Ali, Daniel; Nivievskyi, Oleg
Відкритий доступ
Cloud Platforms and Technologies for Big Satellite Data Processing
(Springer, 2023) Kussul, Nataliia; Shelestov, Andrii; Yailymov, Bohdan
This paper addresses the problem of processing large volumes of satellite data and compares different cloud platforms for potential solutions. Existing cloud platforms like Google Earth Engine, Amazon Web Services (AWS), and CREODIAS have been used to tackle this challenge. However, this study proposes an optimal pipeline for satellite data processing, taking into account the advantages and limitations of each platform. The specific focus is on solving machine learning problems using satellite data. In the experiment conducted, the effectiveness of each cloud platform was analyzed. It was found that cloud platforms offer benefits such as flexibility, access to computing resources, and parallel processing architectures, leading to increased productivity and cost reduction. CREODIAS, in particular, stands out due to its specialization in satellite data and easy access to various data types, along with tools for data searching and visualization. The experiment demonstrated that tasks, from data loading to classification, were executed fastest on CREODIAS resources. However, AWS performed data classification faster. The availability of its own internal data bucket was a significant advantage of CREODIAS, especially when considering ARD data. These findings contribute to the advancement of AI methodologies and have practical implications for solving satellite monitoring applications.
Відкритий доступ
Comparative analysis of classification techniques for topic-based biomedical literature categorisation
(Frontiers Media S.A., 2023) Stepanov, Ihor; Ivasiuk, Arsentii; Yavorskyi, Oleksandr; Frolova, Alina
Introduction: Scientific articles serve as vital sources of biomedical information, but with the yearly growth in publication volume, processing such vast amounts of information has become increasingly challenging. This difficulty is particularly pronounced when it requires the expertise of highly qualified professionals. Our research focused on the domain-specific articles classification to determine whether they contain information about drug-induced liver injury (DILI). DILI is a clinically significant condition and one of the reasons for drug registration failures. The rapid and accurate identification of drugs that may cause such conditions can prevent side effects in millions of patients. Methods: Developing a text classification method can help regulators, such as the FDA, much faster at a massive scale identify facts of potential DILI of concrete drugs. In our study, we compared several text classification methodologies, including transformers, LSTMs, information theory, and statistics-based methods. We devised a simple and interpretable text classification method that is as fast as Naïve Bayes while delivering superior performance for topic-oriented text categorisation. Moreover, we revisited techniques and methodologies to handle the imbalance of the data. Results: Transformers achieve the best results in cases if the distribution of classes and semantics of test data matches the training set. But in cases of imbalanced data, simple statistical-information theory-based models can surpass complex transformers, bringing more interpretable results that are so important for the biomedical domain. As our results show, neural networks can achieve better results if they are pre-trained on domain-specific data, and the loss function was designed to reflect the class distribution. Discussion: Overall, transformers are powerful architecture, however, in certain cases, such as topic classification, its usage can be redundant and simple statistical approaches can achieve compatible results while being much faster and explainable. However, we see potential in combining results from both worlds. Development of new neural network architectures, loss functions and training procedures that bring stability to unbalanced data is a promising topic of development.
Відкритий доступ
Complex method for land degradation estimation
(IOP Conf. Series: Earth and Environmental Science, 2023-01) Kussul, Nataliia; Shumilo, Leonid; Yailymova, Hanna; Shelestov, Andrii; Krasilnikova, Tetiana

Переглянути

Перегляд Кафедра математичного моделювання та аналізу даних (ММАД) за Назва

Результатів на сторінці

Налаштування сортування