Abstract
Classical methods for satellite image analysis appear inadequate for the current bulky data flow. Thus, makingthe interpretation of such images automatic becomes crucial for the analysis and management of phenomenachanging in time and space, observable by satellite. Consequently, this work aims to contribute to the dyna-mic land cover cartography from satellite images, by expressive and easily interpretable mechanisms, and byexplicitly taking into account structural aspects of geographic information. It is part of the object-based imageanalysis framework, and assumes that it is possible to extract useful contextual knowledge from existing maps.Thus, a supervised parameterization method of an image segmentation algorithm is proposed, taking a seg-mentation derived from a land cover map as reference. Secondly, a supervised classification of geographicalobjects is presented. It combines machine learning by Inductive Logic Programming and the Multi-class RuleSet Intersection approach. Finally, prediction confidence indexes are defined to assist interpretation. These ap-proaches are applied to the French Guiana coastline cartography. The results demonstrate the feasibility ofthe segmentation parameterization, but also its variability as a function of the reference map classes and ofthe input data. Nevertheless, methodological developments allow to consider an operational implementation ofsuch an approach. The results concerning the object supervised classification show that it is possible to induceexpressive classification rules that convey consistent and structural information in a given application contextand lead to reliable predictions, with overall accuracy and Kappa values equal to, respectively, 84.6% and 0.7.In conclusion, this work contributes to the automation of the dynamic cartography from remotely sensed imagesand proposes original and promising perspectives.