In this paper, we propose a structural approach to recognizing incomplete graphic objects, which aims to recognize a graphic object before the user finishes drawing all its strokes. In our approach, we first create a Spatial Division Tree (SDT) for the input object and then search for similar objects in the database of graphic object models based on this representation. Along with the SDT, an intersection-based relationship model is also proposed in our approach to describe the relationships among strokes. The experiment result has shown that our approach can significantly save user's input time for complex graphic objects. Moreover, it also proves that our approach can response to user's input in real-time and is independent of the user's drawing style.