The paper describes a method  for generic object recognition , using  
a multi-layered rule-based system. The  representation of 3D objects 
is accomplished with fuzzy linguistic modeling. The objects to be  recognized 
are composed of  3D volumetric  primitives. As inputs for the generic object 
recognition we consider the  contour segments obtained after the edge detection 
and maximum curvature points determination. The information is processed in two stages: 
1) the transformation of existing information into a semantic description, and 
2) generation of production rules which provide the  syntactic relation between 
the semantic expressions. The object recognition in the proposed method are done 
by  aspects  recognition.  Aspect recognition  are performed  by labeling his faces, 
taking into account local neighborhoods.