Mining image time series

Kurzfassung

The process of searching and analyzing data in order to discover potentially useful information is a crucial matter when dealing with large databases. Considering the huge amount of data collected by satellite observation systems, opportunities to generate multispectral image time-series are increasing. Exploratory methods are needed to understand the dynamics of Earth observation scenes and of their objects. In this paper, information mining methods are proposed. They first rely on a hierarchical model based information representation. Entropic measurement similarity in a modelled dynamic feature space characterizing the dynamics of objects and of image structures are then researched