one class CLassification for automatic reading of arabic forms

Abstract

One of the important theses of research at the present time is for computer systems to get rid of the conventional input devices (such as keyboards). This issue is closely related to the language with which the system works. There are outstanding results and systems that work with languages of the developed countries, such as English language and Japanese. Regarding Arabic language, further efforts in the field of research are still to be rendered in order to ripen. Forms and applications data entry is one of the important applications in this field. Therefore we chose in this thesis to apply the idea of one class classification for this issue to have an access to systems that can machine read forms in Arabic Language One Class Classifier (OCC) method is currently receiving much interest. This thesis aims at using “One Class Classifier” method in reading Arabic handwritten forms. The study involved entry of 3500 exit visa forms with scanner and carrying out of all necessary preprocessing (segmentation, noise removal, binarization). Then, a large number of contemporary methods of one class classifier were experimented to read a certain field. The thesis did not get to the required effective system, but it has been a significant step towards such aim. Large datasets were built that may enable other researchers to proceed and develop effective systems. The thesis, also, tested several one class classification- based classifiers, and presented conclusions in a scientific method. The thesis gave recommendations that may help develop these conclusions to achieve the required effective systems.