Unsupervised Image Classification Algorithms Applied to Fire-Prone Area Detection

• Type: Article in International Conference Proceedings Book
• Date: 2025

Authors

Rahimi, I (Author) Other; Lia Duarte (Author) FCUP; Ana Teodoro (Author) FCUP

Conference proceedings International

Title: Proceedings of the 11th International Conference on Geographical Information Systems Theory, Applications and Management

Pages: 136-141

11th International Conference on Geographical Information Systems Theory, Applications and Management, GISTAM 2025

Porto, 1 April 2025 through 3 April 2025

Indexing

Scopus - 0 Citations

Other information

Authenticus ID: P-018-QQS

DOI: 10.5220/0013201800003935

Abstract (EN): Remote sensing data has become critical in identifying fire-prone areas, providing essential insights through satellite imagery and various geospatial inputs. These data sources allow for real-time monitoring, mapping fire susceptibility, and assessing factors such as vegetation, fuel moisture, land use, and environmental conditions. Numerous supervised and unsupervised models combined with remote sensing data have shown great potential in predicting fire-prone regions, offering accurate and timely information for early warning systems and resource allocation. This study focuses on applying two unsupervised methods-PCA, and K-means-using inputs like Sentinel-2 imagery, elevation, and the Zagros Grass Index (ZGI) to identify fire-prone areas in the Kurdo-Zagrosian forests, an area increasingly vulnerable to wildfires. Among the two methods evaluated, PCA demonstrated superior performance in predicting fire-susceptible areas, accurately classifying 80% of the burned regions from 2021 to 2023 as moderate to high-risk zones. © 2025 by SCITEPRESS - Science and Technology Publications, Lda.

Language: English

Type (Professor's evaluation): Scientific

No. of pages: 5