Дисертації | Dissertations

The dissertation is devoted to the development of models, methods, and tools for data processing aimed at monitoring crop yields and integrating with Geographic Information Systems (GIS). As the global population grows, so does the need for food. Efficient agriculture can produce more food per unit of land, ensuring food security for an increasing number of people. Yield monitoring in the context of project management in agriculture is crucial for enhancing its efficiency. This process allows farmers to analyze the impact of various agronomic factors, such as soil type, fertilizer use, and water availability, on crop yields, facilitating resource optimization and reducing environmental impact. Moreover, monitoring is a critical tool for developing effective strategies for managing agricultural projects. Project management is becoming increasingly widespread in agriculture, as it promotes effective organization and management of agricultural projects, enhancing their productivity and profitability. In this context, yield monitoring serves as a fundamental tool for project management, providing valuable information for making informed decisions, planning, and controlling the execution of agricultural projects. Identifying the impact of external and internal factors on yield allows project managers to adapt strategies and optimize resources, ensuring the resilience and efficiency of agricultural initiatives. Thus, integrating project management in agriculture, supported by effective monitoring, opens new opportunities for enhancing productivity, adapting to climate change, fostering innovation, and achieving global food security. Therefore, scientific research on yield monitoring not only contributes to improving the productivity and efficiency of agricultural practices but also plays a key role in ensuring food security, sustainable development, and economic well-being on a global scale. This work addresses a critical task: the development of information technology that includes mathematical models, methods, and procedures for yield monitoring based on geoinformation data (scientific component), as well as the development of a yield monitoring information system that enables the automation of data collection, processing, and utilization of geoinformation data for yield forecasting (practical component). The object of the study is yield monitoring. The subject of the study is models, methods and information technology of yield monitoring based on interaction with Geographic Information Systems. The study aim is to develop models, methods and data processing procedures necessary for yield monitoring. Research Methods. The conducted research is based on methods of systems analysis, technical analysis, artificial intelligence, big data processing, and object-oriented programming. Scientific novelty of the obtained results: • For the first time, an integration model of artificial intelligence for yield monitoring has been developed, based on the combination of multispectral images and geoinformation data. This model integrates Convolutional Neural Networks and Recurrent Neural Networks to enhance the accuracy and sensitivity of monitoring. • The mathematical model of the relationship between phenological indicators and crop yields has been improved. Unlike other models, this combined model includes an adaptive threshold method for determining the membership of crop pixels and identifying the trend and seasonal components of the phenological indicators' time series. This improvement enhances the accuracy of forecasting. • The information technology for yield monitoring based on geoinformation data has been improved. The enhancement involves the use of a combined model of the relationship between phenological indicators and crop yields, as well as an integration model of artificial intelligence. Unlike other technologies, the developed technology takes into account a wider variety of data, which simplifies integration with Geographic Information Systems (GIS). • The methods for representing and storing geoinformation data have been further developed in terms of correlating key properties of agricultural objects with aerospace images of the locality. • Information technologies for project management have further developed in the aspects of monitoring and forecasting yields and integration with geographic information systems. The first chapter, an analysis of the scientific literature was conducted, which established that the use of digital images of geographical areas and the development of Geographic Information Systems (GIS) are becoming key in modern agriculture. This facilitates effective management of cultivated areas, analysis, and prediction of yields, especially in the context of the increasing demand for food against the backdrop of a growing global population. The advancement of technologies provides new opportunities for intensification and optimization in the agricultural sector. It was also established that neural networks are effective tools for yield prediction. They can model complex nonlinear dependencies in agronomic data and process satellite imagery and remote sensing data. It was discovered that no existing service or software combines all the necessary capabilities for crop area management: anomaly detection, identification of phenological changes, and yield estimation. It is shown that a crucial task is to create specialized software that would allow uploading and working with large archives of images and would have built-in methods for intelligent data processing and pattern recognition. The second chapter describes the use of aerospace imagery and time series analysis of images to determine phenological indicators and other important growth and health indicators of plants. The importance of using the NDVI index as a key indicator for assessing plant cover and yield is emphasized. An approach to data processing and analysis within the context of Geographic Information Systems is considered. The vast volume of geographical data requires reduction for computational processing. The chapter also covers the structure and organization of geodata in vector and raster formats, revealing their unique capabilities and limitations for representing and analyzing geographic information models. A conceptual model of the GIS for agricultural monitoring was developed. The system development is divided into four stages: defining objectives, describing functionality, implementation, and diagnostics. Each stage includes steps that facilitate the creation of an effective system for monitoring and managing agricultural crop yields. The importance of a systematic approach to the creation and use of Geographic Information Systems in agriculture is established. The third chapter describes a mathematical model of the relationship between phenological indicators and the yields of agricultural crops and biomonitoring, which considers multispectral field images for dynamic yield forecasting. The decomposition of phenological indicators into trend, seasonal, and random components is aimed at effective yield monitoring. The model includes image binarization steps to define crop areas using a threshold function and the Otsu method for selecting the optimal threshold value. The process of creating and training a hybrid neural network, integrating image data and soil information for yield prediction, is described. The network architecture includes convolutional neural networks (CNN) for image processing and fully connected layers for soil data analysis. This integration allows the network to consider diverse information, enhancing its ability to accurately predict yields. In the second phase, the network uses recurrent neural networks to analyze data sequences, adding the ability to account for temporal dependencies and context. The fourth chapter describes the development of a GIS-based yield monitoring information system to enhance the efficiency of the agricultural sector. The system's modular structure includes modules for data collection, storage, processing, visualization, and analysis, including the use of machine learning for forecasting and process optimization. An algorithm for implementing an artificial intelligence integration model for yield monitoring based on the combination of multispectral images and geoinformation data is described, which includes seven stages: data collection and preparation, neural network development, testing and validation, optimization, implementation in agricultural systems, and further analysis of results. Practical significance of the obtained results. The main scientific provisions of the dissertation have been elevated to the level of methodological generalizations and applied tools, enabling yield monitoring. The agricultural crop yield monitoring information system was validated through comparative analysis methods. Comparing the yield predictions for winter wheat, corn, and barley in the Chernihiv region for 2019 with the predictions using the WOFOST simulation model and data from the State Statistics Service of Ukraine for 2019 shows that the monitoring model can provide sufficiently accurate yield forecasts. It was established that yield is significantly determined by plant development in the first three months after emergence, highlighting the importance of monitoring during this period. The obtained practical results emphasize the potential and limitations of using yield monitoring information technology with Geographic Information Systems. The main provisions and results of the research have been implemented and applied in the activities of Yancheng Polytechnic College.