Open AccessCharacters recognition using keys points and convolutional neural network
Author(s) -
M. Boutounte,
Youssef Ouadid
Publication year - 2021
Publication title -
indonesian journal of electrical engineering and computer science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.241
H-Index - 17
eISSN - 2502-4760
pISSN - 2502-4752
DOI - 10.11591/ijeecs.v22.i3.pp1629-1634
Subject(s) - convolutional neural network , normalization (sociology) , computer science , pattern recognition (psychology) , preprocessor , artificial intelligence , optical character recognition , character recognition , classifier (uml) , feature extraction , artificial neural network , neocognitron , image (mathematics) , sociology , anthropology
In this paper, the convolutional neural network (CNN) is used in order to design an efficient optical character recognition (OCR) system for the Tifinagh characters. indeed, this approach has proved a greater efficiency by giving an accuracy of 99%, this approach based in keys points detection using Harris corner method, the detected points are automatically added to the original image to create a new database compared to the basic method that use directly the database after a preprocessing step consisting on normalization and thinning the characters. Using this method, we can benefit from the power of the convolutional neural network as classifier in image that has already the feature. The test was performed on the Moroccan Royal Institute of Amazigh Culture (IRCAM) database composed of 33000 characters of different size and style what present the difficulty, the keys points are the same in the printed and handwritten characters so this method can be apply in both type with some modifications.