Applying an anomaly-detection algorithm for short-term land use and land cover change detection using time-series SAR images

Abstract

In this study, short-term land use and land cover (LULC) changes caused by human activity were considered as spatial-temporal abnormalities in time-series images. A density-based anomaly detection (DBAD) algorithm was designed to detect the changes. Then the algorithm was applied to RADARSAT time-series images, and synchronous field surveying was performed for validation. The results showed that the DBAD algorithm was good at detecting in-progress construction and newly builtup parcels, with an error of less than 13.3%. A lower detection error was achieved for woodland areas, and a larger error for built-up areas and for some mixed-use land parcels due to the complexity of the parcels.