Abstract :
The objective of the work discussed consists of extracting all components (text, symbols, and connecting lines), constituting the schematic diagram, and recognizing them. For that, the author proposes a system which consists, in a first step, of building hypotheses and, in a second one, of validating these hypotheses by symbol recognition. The author explains the hypothesis building, which performs two tasks: the detection of connecting lines guided by the result of the recognition step and the selection of an area, which is likely to contain a symbol, based on the endpoints of connecting lines. In addition, the author uses not only loops to guide the selection of this area but also the presence of the text which makes it possible to locate symbols not containing loops. The main originality of the approach consists in the fact that it makes intensive use of the cooperation between the detection of connecting lines and the recognition of symbols. The two most important advantages of the use of this strong interdependence are the efficiency of the connecting line detection and the reduction of the number of false loops to be handled
