IOSR Journal of Computer Engineering (IOSR-JCE) e-ISSN: 2278-0661,p-ISSN: 2278-8727, Volume 20, Issue 3, Ver. I (May. - June. 2018), PP 44-59 www.iosrjournals.org DOI: 10.9790/0661-2003014459 www.iosrjournals.org 44 | Page A Novel Recognition of Indian Bank Cheque Names Using Binary Pattern and Feed Forward Neural Network Raghavendra SP 1 , Ajit Danti 2 1 Research Scholar, NES Research Foundation JNNCE College, SHIMOGA, Karnataka, India Email : raghusp.bdvt@gmail.com 2 ResearchHead, NES Research Foundation JNNCE College, SHIMOGA, Karnataka, India Email : ajitdanti@yahoo.com Corresponding Author : Raghavendra SP Abstract: Automatic processing of bank cheques is getting more popularity and attracting more researchers. This paper proposes an efficient machine leaning techniques to design and develop an algorithm to read and recognize the nationalized Indian bank cheques using optical character recognition technique, in which binary patterns extracted by applying classification level decision using feed forward artificial Neural network(NN). In the proposed methodology the neural network is trained to classify six standard nationalized Indian bank cheques. The accuracy of the system is analyzed by the invariant features such as different font, size and varieties of characters with noise and the experimental results reveal satisfactory result. Keywords: Binary pattern; Bank cheque; Neural network; Optical character recognition --------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 14-05-2018 Date of acceptance: 30-05-2018 --------------------------------------------------------------------------------------------------------------------------------------- I. Introduction As the population is growing, bank transactions use of huge number of bank cheques daily. A great deal of work has been done on automatic processing of bank cheques like automatic extraction of items, signature verification, recognition of date from the cheque, amount filled by the user etc. The bank name is written always at the top left corner with unique font, any damage in this region will hamper the performance of the proposed system. The proposed approach automatically segments and recognizes the bank name present at top of the bank cheque, with different character pattern. The proposed approach does not necessitate any prior information and require no human intervention. The system performance is quite promising on large dataset of real cheque images. The concept of geometrical patterns of a character in a given cheque image includes: 1. The changes in the characters of the bank name represent the different visual patterns of the category of alphabets, fonts, size and colour of the query cheque. 2. The physical characteristics of bank name of cheque image are identified using unique binary patterns for the recognition of name of the bank. Bank cheques plays a significant role in today’s cashless transaction of any organization. Cheque images contains fields like bank name, date, courtesy amount, legal amount, logo, cheque number and signature which gives authenticity of the cheque image. In day to day financial transactions a larger set of bank cheques belongs to different banks is a challenging task to automatically recognizing and detecting bank names. Many varieties of systems have been developed to perform cheque image recognition. These systems possess some of the common and similar characteristics. In this approach, an own dataset of bank cheque database is constructed to recognize six nationalized Indian bank chques: Sbi, Canara, Axis, Vijaya, Sbm and Union bank of India. Along with the standard database, hundreds of other cheque images from the internet are also considered. The methodology has two stages. • Segmentation of bank name and extraction of characters • Generation of unique binary pattern for each character • Classification and recognition: In this proposed work, the classification and character recognition is done by feed forward Artificial Neural Network as shown in Figure-1.