Needs and applications for data mining in large series of remotely sensed images

Recent years have shown an increase in image availability at decreasing cost, as focus changed from ¿maximum profit¿ to ¿maximum use¿. This puts analysis and mining of time series of images within reach of a wider audience, promoting development of suitable techniques. This paper focuses on four generic types of mining of large series of images: a) presence and location; b) temporal patterns; c) spatio-temporal patterns; d) moving objects. For each type it is described which group of algorithms are used for mining and how uncertainty can be modeled. The four generic types can be used to find adequate algorithms for data mining and to describe uncertainty for new applications. Further developments are to be expected for tracking of fast moving objects, image mining of mixed archives and irregular time steps. Communication tools for uncertainty in image mining, targeted at users outside the geo-information sciences should be further developed.