The Design And Implementation Of Image Processing And Recognition System In Euro Paper Currency Sorter

The technology of paper currency recognition is a typical issue in the field ofrecognition, and application of the technology developed the paper currency sorterhas also been widely used and promoted. Paper currency sorter has count money,currency examines, paper currency and other features, and it play a more and moreimportant role in bank area. How to identify images is a key problem of papercurrency sorter processing system. Because the system’s real-time requirements, theimage processing algorithms in-depth study has great significance.The subject is from Songhua River CF1000euro paper currency sorter systemof Harbin Institute of Information Technology Co.Hui Tong Eddie’s. The project isin collaboration with Harbin Film Factory. The system of paper currency sorter isconsist of image processing and identifying subsystem, control subsystem andcounterfeit detection subsystem. Our system is the image processing and identifyingsubsystem, it uses the techniques of digital image processing to analyze paperimages and identify it, and sort papers finally.This system was been developed and designed aiming at the euro machinesystem. It included the image preprocessing, recognition of the face value of theimage, recognition of the old and new degree of the paper, identification ofincomplete degree of the paper. Image preprocessing of the paper included threefunctional modules: the brightness compensation of the image, tilt correction of theimage, the edge detection of the image. Recognition of the face value of the imageincluded the selection and extraction of the image characteristics and classifierdesign.For the above questions, this system is mainly adopts the following methods:In the Image acquisition and preprocessing stage, the system adopted theprocessing method for the compensation and the image acquisition synchronous,which reduced time consumption for the pretreatment stage. When the old and newimage is judged, the histogram and blank piece of combination of old and newidentification method was used, improving the stability of the new and olddiscrimination. In the incomplete identification of the image, incomplete area andgrey value of similar background region method was adopted. In the featureextraction, we used the adaptive grid extracted feature method, which effectively solved the problem of the image distortion. In the choice of classifier, we chose theBP3classifier combined with distance classifier method, greatly improving thesystem identification accuracy.The system has been used in the CF1000euro paper currency sorter, the finaltest results fully meet the system requirements. The correct identification upon therate of99.8%, to clear the old error of±5points, the defect error within±1.