Query-by-Example Retrieval in Satellite Image Time Series

The technological evolution of remote sensing sensors led to the acquisition of huge archives of data that are difficult to interpret by human experts. In order to process this huge amount of data and the large number of potential temporal evolutions, multitemporal analysis techniques need to be developed. In this paper, we propose a simple, unsupervised, yet effective technique towards the retrieval of spatio-temporal patterns from satellite image time series (SITS). Following a query-by-example procedure, the proposed method is able to extract patterns that are similar to a given query under two use case scenarios, i.e., short and long SITS, respectively. The experiments prove the successful application of the proposed method in both cases.