Научная визуализация, 2025, том 17, номер 4, страницы 30 - 38, DOI: 10.26583/sv.17.4.04
The Contours Visualization in Satellite Image of Natural Objects by Artificial Intelligence
Авторы: Е.В. Попов1, П.В. Юрченко2
Nizhegorodsky State Architectural and Civil Engineering University
1 ORCID: 0000-0002-3058-2369, popov_eugene@list.ru
2 ORCID: 0009-0009-9619-9125, pavel-yurchenko@list.ru
Аннотация
The paper discusses machine learning methods for satellite image classification. We present a neural network algorithm for visualization of the shapes of natural objects, using a variety of machine learning algorithms for preprocessing the training dataset. Our paper compares the classification of the algorithm, calculates its accuracy, and proposes potential improvements. We tested our approach on satellite images of woodland areas in the Bolsheboldinsky District in Russia. The results demonstrate that our improved neural network algorithm achieves high computational accuracy. Robustness, recall, and overall accuracy reach 0.98, especially using training datasets optimized with a support vector machine (SVM). We also demonstrated the applicability of our method for creating accurate geographic information models and detecting changes in natural resources.
Ключевые слова: geometric modeling, geoinformation modeling, neural network, natural resources, visualization of object boundaries, image classification.
