Thesis title: Integration of SAR and optical data for an automatic land consumption monitoring using Machine Learning algorithm
Urbanization and industrialization, when not carefully planned, can have negative impact on community resulting in deforestation, habitat loss, decreasing biodiversity, etc. Ensuring sustainable urban development and effective planning becomes one of the major necessities for governmental bodies and policymakers. Thus, monitoring changes in land consumption provides crucial insights for decision-makers. This study investigates the possibilities of identifying the changed areas with Google Earth Engine (GEE) which offers a user-friendly platform equipped with geospatial analysis tools like change detection. The study leverages C-band Synthetic Aperture Radar (SAR) imagery from Sentinel-1 and Multispectral Instrument (MSI) imagery from Sentinel-2 to detect changes like new constructions or artificial areas. It combines image subtraction techniques from both MSI and SAR imagery and utilizes various image indices such as the Difference Built-up Index (NDBI), Bare Soil Index (BSI), and Soil-adjusted Vegetation Index (SAVI). Supervised classification using the Random Forest classifier is employed, followed by the application of thresholds to enhance accuracy. Additionally, automatic models for the creation of training dataset and the evaluation of the model accuracy are provided. The study resulted in almost 90% of changes correctly detected for new buildings/constructions and 80% considering all classes.