Paper Currency Number Image Recognition System

Paper currency on the market is the main form of general equivalence. It plays an irreplaceable role in the people's social life. Along with the rapid development of market economy, a lot of social problems have appeared. We can see now mostly malignant crime concerned with currency note,which bring the enormous harm to social security and disturb the harmonious development of society and the people's normal life.The number of paper currency is uniqueness.Any two pieces of real legal never have the same number.According to this properties,at first, registrating numbers of note in the bank,then if supplying the currency market with a note automatic that can identificate number, and identification data will send to a computer, once the same number or not existing,can identify the circulation of money illegal.This paper introduced a kind of design based on ARM and FPGA.With image collection system of non-contact image sensor (CIS) collecting paper currency number,puts forward a kind of high speed real-time data acquisition and preprocessing of image information method.The collected CIS image signal is saved in the SRAM (synchronous random access memory) through pretreatment of occulting output compensation,secondary filter,analog digital converter and binaryzation,getting ready for further image processing by ARM.The driving signal in CIS is designed by FPGA. And the collected image of binaryzation shows clear in the experiment.ARM is responsible for the execution of the program and system image recognition.The collected image signal will be converted image, greyed,binaried,normalized, and other basic algorithm by the system.It was also studied the key algorithms of paper currency recognition system,including:number positioning,character patrition,the geometric method of number tilt correction, the algorithm of number vertical and horizontal slit, the number identification, Neural Network and Artificial Immune System.The experiments show that the collected binary image is clear, the recognition rate reached 10 per second and high reliability of image recognition.This research output can be extended to not only the entry system of securities and bill character , but also other fields'characters identification system.