Circuit Diagrams consist of three basic components: symbols, characters, and connection lines. To Recognize diagrams we have to extract each basic component, and identify their meanings and relations among them. In this thesis, we developed a system for extracting basic components from hand-drawn circuit diagrams. The system first extracts characters by connected component search. Then it generates candidate regions that can be reliably matched to symbol models. Each candidate region is verified by template matching with geometrical information. Finally connection lines are extracted by line analysis.
An experiment is conducted with five sets of four men's hand-drawn diagrams of electronic and logic circuits. One set is used to construct symbol models and training, while the other sets are used for testing. The system achieved 96% accuracy in extracting basic components for the five sets of drawings.