Research On Paper Money Serial Number Identification Based On Genetic Algorithm And Support Vector Machine Theory

Paper money is the basic tool for trading. Each piece of it has its unique serialnumber,while other security information is the same. Financial sector can track and supervisethe cash flow through the establishment the data base of note serial number. Create a paperserial number identification system is of profound significance.The paper is aimed at the study of the paper serial number identification according to theimage characteristics of paper serial number. Using the method of digital image processingand pattern recognition theory knowledge, the author established the whole paper serialnumber identification system,which is composed of the image pretreatment and imagebinarization, character segmentation, feature extraction algorithm and character recognition.The main contents are as follows:(1) Using the genetic algorithm on binary threshold, effectively reduce the binarizationof the chances of getting the interference of image pixels provides a good guarantee forsubsequent image processing.(2)This paper has detailed instructions on the steps of character image segmentation:based on the characters of the image grey value discontinuity and similarity method, separatecharacters through edge detection and image projection, then normalized processing characterthrough the linear processing method.(3) The serial number extraction is through the method of the statisticalcharacteristics,whose accuracy is higher. The statistical characteristics means the pixel pointsvalue within grid through statistical characters.So even after the pretreatment of image thismethod some interference still exists, the method can strengthen the anti-interference abilityof the system.(4) The author chooses support vector machine (SVM) classification method. Theexperimental results prove the quickness and accuracy of recognition of using support vectormachine classifier and the selection of kernel function.