Seeded Classification of Satellite Image Time Series with Lower-Bounded Dynamic Time Warping

Satellite Image Time Series (SITS) record the continuous temporal behavior of land cover types and thus provide a new perspective for finer-grained classification compared with usual spectral spatial information contained in static image. In addition, SITS data is becoming more accessible recent years due to newly launched satellites accumulated historical data. However, lack labeled training samples limits exploration data, especially sophisticated methods. Even straightforward classifier, such as k-nearest neighbor, accuracy efficiency similarity measure also pending problem. this paper, we propose SKNN-LB-DTW, seeded method based on lower-bounded Dynamic Warping (DTW). The word “seeded” indicates that only few are required, not because but our aim explore rich SITS, rather than letting dominate results. We use combination cascading lower bounds early abandoning DTW an accurate yet efficient large scale tasks. experimental results two real datasets demonstrate utility proposed which could become effective solution when amount unlabeled far exceeds