Study On Note Number Automatic Recognition Technique

As a family but special presswork, Notes are widely used in people's everyday life. Owing to the note relate with national revenue, so the note demand high printing quality to keep away kinds of problems in revenue tasks. Number printing is the most important part in note printing, also it is an important means to defend fake. For the moment, note number detection is mainly accomplished by manpower, so it is easy to make mistake, in order to improve detection efficiency, reduce enterprise's production cost, and adapt request of sheet printing machine's high speed and automatic printing, it has an important significance to research note number automatic recognition system.Note printing technique is analyzed, notes which are used in this project are printed by BF41100B sheet printing machine, the note number is printed by NipsonT550, based on the note's character which has lots of line and figure, this paper handle digital image processing technique and adopt VISUALC++ to dispose collected images.Note printing character is analyzed, in the moment of image pretreatment, noise is removed by Gauss template, and image is segmented by iterative threshold, and then carry through morphologic opening operation, choose the given threshold to remove disconnection character. Note number orientation arithmetic is designed based on the character of note, then the note number is cut, and single number is divided and regulated, the gray information of number image's pixel is distilled, in the end, number is recognized by neural network.Experiment indicates that Gauss template has excellent effect. Number orientation arithmetic which based on note character can orient the number area fleetly and accurately, number recognized by the neural network has the merit of high speed and excellent adaptability.