Research On Face Recognition Algorithm Of Paper Currency

In recent years, banknote image recognition technology is a more active subjectin the field of pattern recognition and it has a wide range of applications in thefinancial instrument. Bill Sorter is an electromechanical integrated financialinstrument, which has great market potential. Its main function is to identify thebanknote denomination and face. In this paper, based on the inspection and researchliterature on the basis of previous studies, author makes a study on the image facerecognition in many methods based on the gray characteristics of banknotes.First of all, author makes a simple image preprocessing for banknote imagecollected in advance, including brightness compensation, edge detection and tiltcorrection. Author makes use of some discrete points to fit a straight line in theprocess of edge detection. By detecting edge not only can the author get the edge ofbanknote image, but calculate the center point and tilt angle. The banknote images arerotated at an angle, so that all the images could get a normal position. When the imagepreprocessing is finished, author can get some gray features by meshing a fixedcenter-based area in the banknote image. This method can not only get the mainimage information notes, but has a good versatility to recognize the banknote image.By measuring the size of the bill, the author can identify the denomination of thebanknote. In this paper, the author uses BP network and template matching to identifythe face of banknote image. According to the mesh feature extraction and templatematching using BP network for the identification of the bill. Compared with thetraditional use of artificial face recognition feature extraction, BP network andtemplate matching is more universal, and parallel processing and generalizationability is a good feature of BP network, it is also able to identify some notes whichcontains slight noise and cracks, so the BP network shows good robustness; Templatematching method is simple, which is of strong anti-jamming capability and can beachieved easily in hardware implementation.After some experiments of the banknote image identification under the premiseof rejection, the author finds the recognition rate over99%by the two methodsmentioned, which basically meets the requirements of bill sorter. The methodmentioned in the paper has a good reference for vending systems, automated sortingsystems and financial instruments recognition system.