Monitoring and Detection of Buildings in Forest, Pasture and Highland Areas using Time-Series SAR and Optical Imaging Data: A Case Study of Trabzon

Abstract

The increasing number of Buildings Türkiye’s highlands and pastures, driven by tourism, development initiatives, and zoning amnesty laws, has led to illegal construction, often at the expense of natural landscapes. Despite national laws and international land protection conventions, these regions are experiencing new buildings, threatening traditional land use, particularly livestock farming. This study proposes the use of deep learning models, specifically the YOLO (You Only Look Once) algorithm, to detect building footprints in satellite images. The goal is to enhance monitoring and enforcement efforts, ensuring the protection of highlands, forests, and pastures. To analyse building change trends over the past decade, Sentinel-1 (SAR), enhanced-resolution Sentinel-2 images, and Google Earth satellite data will be utilized. The methodology involves training the deep learning model using highresolution Google Earth imagery and enhanced-resolution Sentinel-2 images, then testing its performance, focusing on the regions most affected by illegal construction. Additionally, Synthetic Aperture Radar (SAR) will be employed to monitor terrain changes. The expected results aim to identify illegal structures, assess the efficacy of fast detection methods for illegal buildings, and provide a foundation for legal enforcement against unsanctioned development.