Collaborative multi-strategical classification for object-oriented image analysis Germain Forestier, C´ edric Wemmert and Pierre Gan¸ carski LSIIT - CNRS - University Louis Pasteur - UMR 7005 Pˆole API, Bd S´ ebastien Brant - 67412 Illkirch, France Email: {forestier,wemmert,gancarski}@lsiit.u-strasbg.fr Abstract. This paper deals with the description of the use of a col- laborative multi-strategy classification applied to image analysis. This system integrates different kinds of unsupervised classification methods and produce for each classifier a result built according to the results of all the other methods. Each method occurrence tries to make its result converge towards the results of the other method occurrences thank to different operators. In this paper we highlight how classifiers collaborate and we present results in the paradigm of object-oriented classification of a VHR remotely sensed image of an urban area. 1 Introduction In the last decade automatic interpretation of remotely sensed images becomes an increasingly active domain. Sensors are now able to get images with a very high spatial resolution (VHR) (i.e. 1 meter resolution) and spectral resolution (up to 100 spectral bands). This increasing precision generates a significant amount of data. These new kinds of data become more and more complex and a chal- lenging task is to design algorithms and systems able to process all data in reasonable time. In these VHR images the abundance of noisy, correlated and irrelevant bands disturb the classical per-pixel classification procedures. In the paradigm of object-oriented classification [1] [2] the image is segmented and the segments are classified using spectral and spatial attributes (e.g shape index, texture...). These new heterogeneous types of data needs specific algorithms to discover relevant groups of objects in order to present interesting information to geographers and geoscientists. Within the framework of our research, we studied in our team many methods of unsupervised classification. All have advantages, but also some limitations which seem sometimes to be complementary. From this study is born the idea to combine, in a multi-strategical classifier [3], many clustering algorithms to make them collaborate in order to propose a single result combining their various results. For that, we define a generic model of an unsupervised classifier. Our system allows several entities of clusterers from this model to collaborate without having to know their type (neural networks, conceptual classifiers, probabilistic clusterer. . . ).