Non-Local Means Filtering based Landslides Coherence Change Detection for Multitemporal SAR Images

Landslides often occur in mountains, where the low coherence characteristics result in decreased accuracy of traditional SAR-based change detection algorithms. In this paper, a non-local means filtering based coherence change detection algorithm is proposed for multitemporal SAR images. First, Persistent Scatterers (PSs) and Distributed Scatterers (DSs) are both detected to improve the monitoring density in low coherence area. Then, coherence change detection is adopted based on the multitemporal SAR images to replace traditional amplitude based change detection algorithm. Finally, non-local means filtering is used to improve the change detection accuracy of the whole area. The raw data of landslides occurred in Hong Kong is used to indicate the effectiveness of the proposed algorithm.