A Sketch Edit System Based On Space Reasoning

Hand-drawn sketch edit is a basic task of sketch understanding, which needs to solve the uncertainty and ambiguity of sketches. It is a necessary part of the sketch understanding.Current sketch understanding systems usually have two states and transform reciprocally by a specific sign. It can't distinguish between the geometry and gesture edit command according to the sketch. Current edits always depend on the traditional menu commands and don't get rid of the bondage of the traditional graphical interfaces.To address this problem, sketch recognition and edit method based on SVM and space reasoning is proposed and a sketch edit system is designed and accomplished, which combines the Support Vector Machine with space reasoning to recognize the sketch and distinguish between the geometry and gesture edit command. A series of gestures are designed and edit can be accomplished dispensing with menu command.The state-of-arts of sketch recognition methods are reviewed first in this thesis. Then sketch classified methods based on SVM theory are discussed and space reasoning theory, which is applied to sketch edit, is analyzed. Preliminary results on UML class sketches understanding demonstrate that this system is feasible and promising.