Remote Sensing and Machine Learning on Anomaly Detection at high spectral and temporal dynamics regions in Brazil

Abstract

In climate changes context Remote Sensing tools are widely used and widespread in research. In this sense, Artificial Intelligence rises offering possible improves for environmental monitoring applications using techniques such as Machine Learning for Anomaly Detection applied to Remote Sensing imagery to identify the spatio-temporal changes over the Earth’s surface. This approach is explored in three high dynamic regions in Brazil assessing deforestation, fires and technological disaster areas using One-Class SVM and Isolation Forest methods over MODIS, Landsat and Sentinel images.