Face Identification From A Single 2D Line Drawing

This thesis is the project supported by Shanxi National Science Foundation (No.2002E224).Line drawing is an important medium for people themselves to exchange 3D information or with computers. Line drawing of complex objects not only includes scenograph, affine graph, linear projective graph and nonlinear projective graph, but also freehand sketch and graphics drew from images. Reconstructing a three-dimensional object from line drawing is the important parts of the computer vision research. Face identification is the key to the 3D reconstruction from a single two-dimensional line drawing. For the result of the face identification is the precondition of much research related to line drawing, we need to do corresponding research on face identification from a single line drawing.In this thesis, our research is based on given basic assumption and precondition. Firstly, we search three-dimensional implied information from the single two-dimensional line drawing. After that, a serial of elementary theorem and interrelated deduction are proposed to make certain real faces as searching cycles and help to automatically delete a set of cycles that are not real faces. It can save run time and memory space. Secondly, we put forward a recursion-based arithmetic for searching for all cycles from a line drawing. And the whole method based on fore-mentioned theorem and deduction is proposed to search for faces from the cycles. This method is applied to objects with planar faces. Finally schemes to deal with objects with curved faces and objects each represented by two or more disjoint graphs are also proposed. And we describe a synthetical method which can be applied to general objects. The method can identify faces from line drawing showing general. Toplanar objects not with holes, recognized result is reliable. Even to objects with curved faces or with holes, recognized result is also better. Comparing with other arithmetic, our method improves more.