Akademik Veri Yönetim Sistemi
SURFACE TOPOGRAPHY-METROLOGY AND PROPERTIES, cilt.13, sa.4, 2025 (SCI-Expanded, Scopus)
Sinkholes are surface depressions or collapses that often form due to the gradual erosion of soluble underground rocks. Their sudden appearance can pose serious risks to landscapes, infrastructure, and communities. Monitoring sinkholes is essential for understanding changes in Earth's surface over time and for mitigating potential hazards. This study explores the use of Artificial Intelligence (AI) to support the detection and monitoring of sinkholes through satellite imagery. A pre-trained DenseNet201 deep learning model was fine-tuned to recognize sinkhole-related features in Planet satellite images captured over years. The model demonstrated efficient performance, achieving 88% accuracy during training and validation. It also identified six additional patches containing sinkholes between 2020 and 2024. When data availability, site coverage, and image resolution are sufficient, these results suggest that well-adapted deep learning models can significantly enhance environmental monitoring and geospatial analysis.