Annotation:
Computerized recognition of fundamental electrical circuits (FES) railway
automation and remote control (RARC) is an urgent and diffi cult task. The decision
can be reasonably divided into the decision of the individual sub-tasks. Thus, the
general recognition algorithm is divided into several specialized algorithms and the
decision becomes more simple and straightforward. The main sub-tasks are selection
and recognition of the FES structure, of the text, of the stamp and of other information.
The article describes an approach for text processing on the FES RARC.
Text information of the FES is extremely important and without its analysis
it is impossible to provide a complete transmission of the data of scanned FES
RARC image into electronic form. The article proposes an algorithm for selection
of text information of FES, using the clustering algorithm to select groups of
symbols, as well as method of preparation of unique separated expressions (lexical
units) within the group.
The article also describes the problem of analysis of obtained versions of text
expressions and methods of selecting the most correct options. At the end of the
article there is a general fl ow chart of the FES processing method, taking into account
the methods of analysis of textual information proposed in the article.
Description of the study ends with the development of a software prototype,
that implements the methods, mentioned in the article, and its using on a test sample of 300 printed and then scanned FES of varied quality. The paper also
provides the conclusions.
Key words:
technical documentation; image recognition; text recognition; elementary electric
diagrams