Summary: The form features recognition problem was often studied as a problem highly coupled with the manufacturing processes (generally machining processes). Therefore, this approach is not well suited to the new modeling environment based on concurrent engineering, where different designers cooperate for the product definition. The paper presents an approach which considers these new contexts by splitting this complex problem into two subproblems. The first one is purely geometric: it detects the elementary shape characteristics that enable to code the part in a context independant language. The second one would read this language under semantic rules depending on the viewpoint. The first step, presented in the paper, is based on a morphological and topological analysis which generates a set of shape features connected by geometric constraints.