This work is designed to produce an automatic tool to update data bases used to build geographic maps. The main idea is to perform a cooperative process between the most efficient, fast and robust algorithms; some of them exist and need to be adapted to get the best.

The early results show that the global principle is good:

• the snakes used as complement to dynamic programming are really efficient, robust and fast;
• the modularity of the concept allows adapting the system to different types of images, by adding new operators and rules to manage the extraction and the fusion.

The addition of new operators and pooling fusion should yield the whole process more reliable. The integration of measures to evaluate the risk of confusion may be useful to qualify the results and to help the verification of vectors.

                  
  
 IRIT-UPS