Change Detection of Flood Hazard Areas in Multi-Source Heterogeneous Earth Observation Image Time Series Based on Spatiotemporal Enhancement Strategy

Abstract

AbstractChange detection (CD) on multiple remote sensing images has been widely used for monitoring flood changes. In this paper, we innovatively propose a spatiotemporal enhanced CD (STECD) algorithm, which exploits the spatial and temporal dependence of multi-source heterogeneous (MSH) Earth Observation image time series (EOITS). Our STECD algorithm mainly contains two steps, i.e., spatial clustering and temporal enhancement. In the spatial clustering step, we propose a sparse Markov random field (SMRF)-based strategy to iteratively optimize the boundaries of flood areas in accordance with contextually spatial features in each image of MSH EOITS. In the temporal enhancement step, the historical incremental information of flood areas is employed as constraints to effectively reduce the effects of terrain shadows (in synthetic aperture radar (SAR) images) and cloud shadows and topography shadows (in optical images) on spatial clustering results in accordance with the temporal dependence among MSH EOITS. Experiments on real MSH EOITS show that the overall detection accuracy of our proposed STECD algorithm is higher than existing commonly used methods for CD of flood hazard areas.KeywordsSpatiotemporal enhanced change detectionMulti-source remote sensingFlood