Introduction

All approaches concerning pattern recognition within remote sensed images aim to give a realistic description of the corresponding real scene, for example maps production from satellite images.

The up-to-date approaches take into account out-image data and expert knowledge about

• searched classes in a classification process or
• objects in pattern recognition processes.

For example, a human expert who wants to produce a vegetation map from remote sensed images, will in fact manage with three types of information sources:

• the digital images, but also

• available cartographic information or visual information from the image such as topography…,

• his knowledge about usual contexts for the vegetation categories searched.

 So, pattern recognition algorithms have to combine these three types of data and should allow to model and integrate expert knowledge and sometimes know-how.

This may be reached by such techniques as expert systems, or fuzzy neural networks.