A Research For Banknotes Crown Font Recogniton Technology On ATM Machine

Crown word number refers to a combination of a string of English characters and Arabic numerals on the RMB, it has only one unique number of a banknote, which is the special identity of the RMB banknotes. It has never stopped since the beginning of counterfeit money phenomenon, although the anti-counterfeiting technology has never stopped, too. Occasionally it’s heard from the news that some people take out some counterfeit money from ATM machine. In order to prevent criminals create more counterfeit banknotes, the identification technology of banknotes crown font becomes more and more urgent, and it is a worthy topic for deepening of the study to go on.In this dissertation, the RMB crown word number identification technology is based on ATM machines systems. For the RMB’s unique features, the image processing technology has been used, and it’s combined with optical character recognition theory, a set of recognition method based on the crown word characters is created.The first part is image acquisition. The Contact Image Sensor is used for image acquisition in this dissertation, and then the noise is removed from the image. By analyzing the causes of noise generation, mean filter and median filter methods are selected for image filtering, it shows from the experimental results that to remove the noise from the banknote image, median filtering method is the best.The second part is image processing. It has a series of algorithm processing after the image de-noising. To save time and improve efficiency, local threshold method is used to get binary image. The priori knowledge is used to locate the crown size area which is based on the particularity of RMB. Then the Hough transform is used for image tilt correction, and a projection method is used to achieve delimiting character. Finally, a linear normalization method is used to complete the character’s normalization.The third part is feature extraction. A feature extraction method based on eight directions is used in the dissertation by analyzing the structure and statistical characteristics. And the characters’ gradient feature was extracted. At last, the data is saved for when it is convenient to use to identify the next chapter.The fourth part is Character recognition. This thesis describes and analyzes some recognition algorithms, such as template matching identification method, it’s based on eight-way mesh gradient characteristic method, nearest neighbor, Traditional support vector machine method, etc. Respectively, the result is comparative analyzed in theory or results, This thesis proposes a method for identification, it’s based on support vector machine recognition algorithm(SVM) for crown size identification, which is called an improved parallel processing SVM recognition algorithm. It shows from the results that the algorithm can not only meet the ATM system to identify the timely request, but also to meet the recognition accuracy. Basically, it can meet the needs of industrial inspection.In this thesis, the image acquisition system for ATM machine image sensor is selected, and the noise is finally effectively removed. By a series of image processing algorithm, it achieves the image tilt correction, character segmentation, size normalization and extract effective features characters, and it provides effective data for character recognition, and ultimately in the real-time conditions to ensure the system so that the banknotes crown famous character recognition rate up to 97% and it’s able to meet the actual recognition demand.