Algorithm Research Of High Speed And Precision Banknotes Sorter

In recent years, banknotes are more active in the field of pattern recognition, and it has a broad application prospect. The banknotes Sorter developed by banknotes recognition technology is playing an increasingly important role in financial systems. In the process of banknotes recognition, the collected images are obtained in the low-resolution. In such conditions, how to make the recognition speed and accuracy meeting the requirements of Sorter is the key issue in the banknotes recognition. Under the analysis of banknotes characteristic, a detailed study about the banknotes recognition is done in this paper. In pre-processing stage, for the allowance of the real time demand, we dispose the method of Hough transform and apply linear fitting to determine the border of paper currency to achieve directly localization and adjustment. In the feature extraction, the method of painting the grid on the Chinese Money Banknote (RMB) is used to extract the rough characteristics. After, Self-organizing Map (SOM) is used to cluster similar banknotes characteristics together, and the Principal Component Analysis (PCA) is used further to extract banknote image feature. In the classifier design, we have put forward three algorithms - Dynamic Time Warping (DTW), the combination of Support Vector Machine and Dynamic Time Warping (SVM / DTW) and Counter Propagation Net (CPN). Three methods are used to recognize banknotes, and the results show there are Respective advantages and disadvantages in them. The speed of DTW is faster, but the accuracy rate is not high; the recognition accuracy rate of SVM / DTW is higher, but it can only recognize the old and new grades of banknotes and cannot recognize the direction of banknotes. The recognition speed and accuracy of CPN are high, and the direction of banknotes is recognized by it. Experiments show CPN algorithm meets the requirements of banknotes Sorter, and its recognition performance is greatly improved.