Automatic Identification Of Large Part Of RMB Image Features

With the rapid development of economy in our country, the number of the market circulation of paper currency is growing. But the existence and spread of counterfeit banknotes seriously interfere the country’s economic order. In the face of the new fraud means and technology, using image processing technology to identify the counterfeit banknotes draws more and more attention. Through the image processing technology, the paper currency can be accurately identified and classified, so it can guarantee the safety of paper currency in the circulation process.Based on the algorithm of the image processing and pattern recognition, in this paper, a research on the identification and recognition problems under the transmission light image and infrared image of the Y 100 has been studied. It respectively introduces the authentication on the white watermark image and the transmission light image of optically variable ink(OVI) acquired by CMOS, and on the infrared image acquired by the Contact Image Sensor(CIS). The specific tasks include the following sections.This article firstly introduces how to obtain the white watermark region image and OVI region image, including the pretreatment process such as the barrel distortion correction.Secondly, cut out white watermark image and OVI image according to the positioning information. In order to extract features from the white watermark image, three algorithms have been adopted, including a new NCC algorithm based on the projection difference、an algorithm based on the connected domain and eight-directions feature extraction algorithm. The analysis and comparison to these three algorithms also have been carried on. In order to extract features of the OVI image, the concept of color temperature and color aberration have been introduced. The color aberration is taken as a part of features, and finally a three-dimensional feature is obtained. Then the KNN and Perceptron algorithm are designed as classifiers and realize the authentication function.Finally, it introduces the counterfeit detection process on the infrared image acquired by the CIS, including the acquisition of the basic information and authentication of banknotes. In order to extract features, a new method through extracting the relative gray level information from the feature areas comparing to the adjacent areas as features is put forward. And the information of security thread in the infrared image can also be used to authenticate the banknotes.A lot of experiments confirmed that the respective rate of identifying the counterfeit banknotes of using the features of white watermark and OVI image can be more than 90% and 80%, and a very good result also has been achieved by the algorithm under the infrared images.