Primitive Recognition on the Base of Combined Relationships between Primitives
Yue Xiao · Chinese Journal of Computers · 2002
Primitive recognition is a key and difficult problem in 3D reconstruction. Some different techniques used in primitive recognition include bottom up approach, pattern recognition, rule based methods, etc.. To date, they are still far from the desired goal of a general solution. This paper presents a new approach to the problem. The main underlying philosophy is to view a complex part as being composed of primitives with different kinds of combined relationships among them, and to recognize these primitives by making use of the knowledge about the property of their 2D representations. Firstly, the combined relationships among the different primitives are classified into two kinds, namely, explicitly and implicitly combined relationship. Secondly, pointed to the 2D projection character of explicitly combined relationship, their corresponded projection in 2D drawing are defined as combined factor which was taken as the basis of combine relationship table. Then the primitives with different kind of combined relationships could be recognized on the base of combined relationship table. Because the technique establishes the connection between the 2D combined factors and their homologues 3D combined relationships, it can limit the search space and extend the recognition shape. Complete details are provided to illustrate primitive recognition using this approach.