Study Of Recognition For Hand-drawn Electronic Component Symbol Based On Input With Smart Image Card

CAD is the most extensive and active applied realm that computer graphics used in the industry field. Now the technique and system of CAD have been developed perfectly, which are used at later stage of design, but can't satisfy the demand of concept's design at early stage nicely. Consequently, it is important and realistic to develop new input tool suitable for sketch design and then study sketch recognition and understand in this input way. Electric circuit diagram design is an important and applied realm of CAD, automatic input and recognition of electronic component symbols are studied thoroughly in this paper.It is the first time that Smart image card is used to be a tool to input electronic circuit diagrams. Designer can write sketch directly on it as same as on papers. It accord with men's habits of thinking and writing. Its operation is very simple, convenience, nature, and its process is very fast and accurate. The result of sampling is ideal. Therefore, it is very suitable to be used as the input device of electric circuit diagram.Preprocessing is the important part of handwritten electronic component symbols recognition. Preprocessings such as smoothing,thinning,normalization are done on primitive image of handwritten electronic component symbols. Normalization is the valid path to resolve transform of image, so several kinds of normalization algorithm are experimented to select better one suitable for recognition of handwritten electronic component symbol.Feature extraction is key technique of pattern recognition. this paper introduced the important application of feature extraction in the process of symbols recognition. According to the characteristic of handwritten electronic component symbols, both comparative feature based on expandable meshing and primitive shape distribution feature based on expandable meshing are presented, whose application value are analyzed thought recognition experiment.Design of classifier is another key technique of pattern recognition. integrated technique of mutilclassifier is analyzed, and a model of symbol recognition with two class classifier is proposed in which input feature data is comparative feature based on expandable meshing with distance classifier at first class and input feature data is primitive shape distribution feature based on expandable meshing with BP neural network classifier at second class. Decision strategy is presented, then a recognition system is designed to verify the validity of the proposed methods.