Paper Currency Image Recognition Technology Research

Along with our country economy rapid development, the scale of currency issue and circulation become bigger and bigger, the domestic requirement becomes very urgent for financial equipments such as bank note sorting equipment. At the same time, along with informatization, automation degree are continuously improvement, the smart devices such as the vending machine, automatic deposit machine, the ticket machine are entering into more and more fields. The image recognition of paper money is a necessary function module. This paper expounds the basic concepts about the image recognition technology of paper banknotes, analyzes and discusses some commonly used thoughts about banknotes image recognition, and the banknotes image recognition technology is studied thoroughly. On the basis of looking up the large numbers of home and abroad reference, aiming at Chinese paper currency’s characteristics, the recognition algorithm of the banknote denomination, face and orientation, authentic and false banknotes is designed and improved. The fifth edition RMB which is extensively circulating in domestic is the researched object. The research work is mainly focused in the following aspects:First preprocess the sampled paper banknotes, mainly includes images gradation, denoising, image segmentation extraction and related steps. This is a necessary work for image recognition in order to better recognize the denomination, face and orientation, true and false.Then through the template matching extract the characteristic block of denomination region. The twin peaks threshold value method is used to segment the numeric part from the denomination characteristic block. Then use the vertical projection to recognize a single character from the numeric part and realize the recognition of banknotes denomination.Next part is the banknote’s face and orientation recognition. According to the feature, that is, the grayscale feature in the fixed region is different greatly, the face and orientation of paper banknote is judged by gray weighting on the extracted feature region.Finally, this paper brings forward a new kind of a paper currency’s authenticity identification method. Because the true and false banknotes texture feature in specified region is differed greatly so that it is difficult to copy, a true and false identification method is put forward by utilizing the texture multi-resolution recognition feature of wavelet transformation on different resolution scale. Experiments show that the authenticity identification method of banknotes given by this paper is reasonable and has some theoretical significance and practical utility.