Research Of The Paper Money Number Recognition System Based On S3C44B0X

1.IntroductionThe robberies of large-denomination paper money influenced the safety andstabilization of the society due to the money with a striking amount. We register itsnumber before the paper money goes into storage. If we find the number is thesame as that of the paper money robbed, we can restrict its currency, and clue thepoliceman to the case according to the uniqueness of the paper money number.Simultaneously the criminal offenders are afraid of that. If find that the numbers ofthe identical face value are the same, we can conclude there are false paper money,which is helpful in the recognition of the super false. As a result of the requirementabove, the paper has designed a recognition system of the paper money numbersbased on S3C44B0X.The recognition of the paper money number is a comprehensive technologyand it is one kind of the optical character recognitions, including the digital imageprocessing, the pattern recognition and so on. Its general recognition process asshown in Figure 1, mainly has following several links:Figure1 Number Recgnition Flow Chart1) The image sensor expression paper money image by certain quantificationgrey level and the resolution, the digital image which produces must be able todemonstrate difference between the character and the background.2) The main purpose of the preprocessing operation is to remove theinformation that can disturb the result of the correct recognition, and reserves thecharacteristic information of the character as much as possible, it mainly includes:noise reduction, the binary conversion, the number region localization, thegeometry slope and the grey level adjusts, character segmentation and so on. Whenthe processing request different, the order of each preprocessing approach isdifferent possibly, and some methods possibly need to be used several times, forinstance,the image is reduced the noise at the begin and end of the binaryconversion.3) The purpose of the characteristic extraction is to extract the uniquecharacteristic (one kind of or the combination of several kind of characteristic) inthe character set, then the recognition of character has uniqueness. The charactercharacteristic is determined by the algorithm request, and the algorithm differentpossibly needs characteristic different.4) The end processing is to enhance the correct rate of character recognition.2.Research contentThe main purpose of the paper is to develop a recognition system of thepaper-money number, and achieves the acquisition of the paper-money image inhigh speed and real-time processing of the image, and extract the numberaccurately, its main research content is divided into two parts:The first part provides a fast algorithm that is suitable for the paper-moneyimage preprocessing and character extraction.In order to reduce the data amount in the character recognition, this systemadopts an algorithm based on the progressive gray value to realize the numberregion localization of the paper money.Because the histogram of the number region has no obvious double-crestwaves, it can not get the binary value threshold by the way of valley value. Thepaper adopts specific steps of an iterative method to get the best binary value,which segment character from background better.The paper puts forward an ergodic method to find out the highest point, andadopts the Least Mean Square (LMS) to get the straight line, and finds outthe lean of the number region, and at last uses the near interpolation method torealize revolving to adjust the image, Suppose there are some vertical lines thathave no intersection with the characters, these vertical lines separate the charactersfrom each other.The characters of the paper money include 26 English characters and 10 Arabicfigures .These total 36 kinds of characters can be classified into two kinds, and thefirst two characters and the last 8 figures on the location of the character. We canuse different decision trees to distinguish them first and then recognize them, whichcan avoid the disturbance between the character and figure feature information.The paper adopts a method completely based on the structure knowledge of thecharacters, which takes the structural characteristics, the stroke type, the quantity,the location and the intersections of the characters and imaginary lines as therecognition characteristics to generate decision tree. The decision tree can be usedto recognize the characters and the figures in different group.The second part introduced the hardware composition principle of the papermoney number recognition system. It is mainly composed by the following parts:(1) Image acquisition module: the module realizes the acquisition andquantification of the analog image information, generates digital image data. Theoutput of the digital image data and control signals can be set according to adefinite format. The module uses the digital CMOS image sensor OV7620 (madeby the Omnivision Corporation of American) to receive the control command ofthe processor, and puts out digital image data and control signal according to acertain format.(2) Generation unit module of the read-write and control signal: the modulerealizes the generation of the read-write and each kind of control signal, includingthe control signal of the storage for the digital image data, the read signal that theprocessor operates on the FIFO, the handshake between the FIFO and the processorand the reset signal for the FIFO.(3) Data buffer unit module: the module realizes the parallel storage and readof the data, which enhances the parallelism of the system and saves the time for theprocessor waiting for the image data. The module uses Averlogic CorporationAL422B as the FIFO to complete the high-speed read and storage.(4) The processor and periphery circuit module: the module is composed by thefollowing several parts, the processor CS3C44B0X, the clock reset circuit,SDRAM, Flash ROM and serial port. The processor S3CS3C44B0X takes chargeof the pro and con direction localization, the noise reduction, the binary conversionand the characteristic extraction. It uploads the number information of the papermoney to PC and accepts the control command from PC using serial port .In thisway, it realizes the communication with PC. The Flash ROM SST39VF160 is usedto stores the code and the data when it is not electrified. SDRAM HY57V1620 isused to stores the data area, the stack area and the running program code, it canenhance the perform speed of the program.3. ConclusionThe system has realized the image's high speed acquisition by the hardware andfast the paper-money-number recognition by the software and basically achievedthe predetermined target.