Research On Image Recognition Algorithm Of RMB Number

RMB, as the legal tender of our country, is of obvious significance, so it is a pressing issue to effectively supervise RMB. According to the uniqueness of RMB, financial sector builds a number profile before the storage of RMB and make advantage of OCR(Optical Character Recognition) to realize the fast recognition number, not only completing paper currency tracking but also counterfeit money identifying, which plays a good regulatory role on currency circulation.The paper, taking the fifth edition one hundred yuan RMB as the research object, applying relevant knowledge of image processing and pattern recognition on account of the characteristics of RMB number image, realized number image preprocessing including Image Filtering Denoising and Binarization, Image Correction and Positioning, String Localization Method, Character Segment and Character Size Normalization. At last, realizing the image recognition by choosing a kind of recognition algorithm based on three architectural feature of characters.An ARM based embedded system is used as experiment platform, realizing the whole experiment process from collecting image to identifying results, elaborating on the problems such as characters and fonts differences, characters interference in the experiments and proposing solutions. Through analyzing and comparing on high identification of several kinds of recognition procedure, we can see that the algorithm in the paper has high identification and can well meet actual demand.