Satellite Image Time Series Clustering via Time Adaptive Optimal Transport

Satellite Image Time Series (SITS) have become more accessible in recent years and SITS analysis has attracted increasing research interest. Given that labeled training samples are time effort consuming to acquire, clustering or unsupervised methods need be developed. Similarity measure is critical for clustering, however, currently established represented by Dynamic Warping (DTW) still exhibit several issues when coping with SITS, such as pathological alignment, sensitivity spike noise, limitation on capacity. In this paper, we introduce a new series similarity method named adaptive optimal transport (TAOT) the application of clustering. TAOT inherits promising properties comparing series. Statistical visual results two real datasets different settings demonstrate can effectively alleviate DTW further improve accuracy. Thus, serve usable tool explore potential precious data.