Machine learning, neural networks

Тesting neural network models in solving project-oriented problems of land analysis

E. O. Ladanova, V. V. Nikulin

National Research Ogarev Mordovia State University, Saransk, Russia

Abstract: This paper discusses the application of machine learning methods, in particular convolutional neural networks, to analyze land plot changes based on satellite images. The relevance of the work is due to the need for effective monitoring of natural resources and changes in land use in the context of increasing urbanization and climate change. The article presents neural network architectures, such as U-Net, ResNet and Siamese networks. The main focus is on the development of a new architecture based on Siamese networks with an attention mechanism, which allows increasing the accuracy of analysis by focusing on key features. The importance of using machine learning methods for analyzing geospatial data is emphasized, opening up opportunities for monitoring changes and effectively managing natural resources.

Keywords: machine learning, convolutional neural networks, satellite imagery analysis, change detection, siamese networks, attention mechanism.

DOI: 10.14357/20718594250205

Bibliographic databases:

•