Go to MathAI 2025 Conference homepageModel for segmentation of forest logging in northern latitudes using Sentinel-2 images based on an ensemble of multi-specialized segmenters
The problem of automated decryption of illegal logging in the forests of the Khanty-Mansiysk Autonomous Okrug-Yugra is considered. Due to the human factor, a share of such areas is missed. We are talking about the impossibility of catching some of the violators who avoid a fine. The option of involving more operators in the work may naturally entail unreasonable financial costs. By applying methods and algorithms of neural network segmentation, we intend to significantly reduce the costs of decrypting changes in forest lands for violations of environmental legislation.
This task was already considered for our district earlier. In this article, we present a solution based on an ensemble of multi-specialized segmentors, which exceeds the mentioned one by 3.1% in F1 score. As an initial data set, we use summer Sentinel-2 satellite images of the Khanty-Mansiysk Autonomous Okrug-Yugra territory for 2018-2022 with manually marked fellings.