Paper Currency Identification And Distinguishing Technology Research In The US Dollars Based On Multi-spectral Images

The US dollar is the most widely used international currency.Anti-counterfeiting technology based on multi-spectral images of the US dollars has been the hot spot in the field of paper currency anti-counterfeiting, which relates to the national financial security The research has important theoretical research value and widespread reality.Multi-spectral images of the US dollars contain the image information of visible light, the infrared reflection image information,infrared transmitting information and UV information.Through analyzing the multi-spectral images of the US dollars, we can reliably apart the true from the fake and be able to ensure the safety and reliability of the US dollars in circulation by finding out the fake money that traditional anti-counterfeiting technology can't identify, as well as the money that are altered and not suitable for circulation.Domestic study of the US dollars is still in its infancy, The traditional banknote image recognition algorithm is not applicable. The main problems of anti-counterfeiting algorithm based on multi-spectral images of the US dollars will be faced as following:Firstly, the USD has many versions and image texture similarity is high,witch makes it difficult to distinguish USD versions. Secondly, Dollar bills of different denominations have the same size,which make it difficult to distinguish the denomination. Thirdly, the US dollars in circulation have different degrees of old and new, wear, pollution, and so on.It raises very high demands during our image pre-processing, denomination and direction recognition, as well as authenticity identification.To solve these problems, multi-spectral images of USD are processed in four steps:pre-processing,directions versions and denominations recognition,True and false recognition and fineness analysis.Firstly, we improve the edge detection algorithm through analyzing the anti-counterfeiting characteristics of the US dollars. Secondly, this paper puts forward towards the direction identification method based on useful picture size, Haar feature and SVM;version identification based on circle detection and gray feature;denomination identification based on regional Gaussian descriptors.Thirdly,this paper uses improved LBP to extract texture feature and uniform LBP to reduce dimension of the gray level histogram,and Chi square distance is used todistinguish the forged paper currency.Lastly,in order to meet the market requirement of distinguish the new and the old paper currency,we proposed Gaussian model to solve this problem.This algorithm solves the problems of weak anti-interference ability, computing workload and low accuracy existing in the traditional algorithm.By comparing with other image algorithms, the experimental results show that the proposed algorithms have characteristics of computing fast,accurate and high robustness.