Mining change patterns in time series high‐resolution SAR images

Abstract

Here, the authors use the K-Shape algorithm, which is based on a z-normalised version of cross-correlation distance measure, to extract change patterns from time series synthetic aperture radar (SAR) images. These change patterns focus only on the shape of time series, rather than amplitude and phase. Besides, they employ a simple and elegant method to choose the optimal number of clusters. Experimental results on multi-temporal SAR images of Shanghai indicate the effectiveness of their approach.