Sustainable Development of Rural Areas: Machine Learning for Semantic Segmentation of Agricultural Fields

Significant growth in the world population is expected over the coming decades, driving up the demand for food. Under these conditions, the role of rural areas and the need for their sustainable development are growing. Global climate change trends are leading to significant fluctuations in rainfall and temperature, requiring the agricultural sector to rethink planning and change the way it works. An integrated approach to soil fertility management is needed to increase the productivity and resilience of agricultural systems and promote food security. To make informed decisions, it is necessary to process and analyze Big Data, including using machine learning. This will provide the necessary information, which is essential for effective decision-making by agribusiness regarding the targeted application of a set of measures to improve soil fertility. Considerable assistance in this can be provided by the analysis of data over a long period of time on the types of crops grown in the fields of a particular farm. To determine the range of cultivated crops on the analyzed lands, it is proposed to carry out semantic segmentation of fields using a computer vision model. The authors created a model of semantic field segmentation based on a convolutional neural network. The chapter presents the results of semantic segmentation of Ukrainian fields on the example of Poltava and Dnipropetrovsk regions, which mainly correspond to statistical data on the areas of fields occupied by various types of agricultural crops.