Research And Application Of Freehand Sketch On-line Recognition System

Rough graphics recognition system assists designers to display design thinking and createcreative works. More and more experts and scholars focus their research on sketch recognitionsystem. On-line identification graphics is a crucial part in intelligent sketch recognition, as abranch of pattern recognition. After a long time of development, attention to its strength hasreached the unprecedented level.The most CAD tools don’t use the traditional paper-pencil method, and the users mustselect pre-defined graphic elements in a lot of buttons and menus. The method requires users totrain with interacting before using the tool, and during the operation, the method is very flexible.At first, the processing of rough graph and the recognition of graph’s type are important inthe research of sketch. By resample, hand-painted discrete touch points of the graph can be got.Then the stroke is smoothed and segmented. The one-dimensional representation of strokes isdesigned. When the stroke is zoomed or rotated, the one-dimensional representation is invariable.By analyzing the structure characteristics of the stroke, a two-stage classification algorithm toidentify strokes is designed. The first stage classification algorithm matches the Euclideandistance of the strokes. The second stage classification algorithm adopts ELM. In addition, therecognizer can recognize and predict the stroke, according to the model of user’s habit based onthe sketch’s context.An on-line recognizer is designed: the hand-drawn sketch system. There are almost nobuttons and menus in the system, and all commands are controlled by the designed strokes.Therefore, in the system, there are two types of stroke: primitive graph strokes and commandstrokes.