Fast Method For Paper Currency Number Recognition On RMB

The number is one of the important features of paper currency, which is not only the identifier for issuance amount of paper currency, but also plays an important role in money circulation statistics, money management and other aspects. Due to the unique nature of each paper currency number and it’s usage to identify the authenticity and identity of the paper currency, the automatic recognition for paper currency number is of wide application value.Today, theoretical knowledge and hardware technology needed for automatic recognition of paper currency number are already fairly mature. This dissertation firstly introduces both hardware and software of the whole automatic recognition system. Then a fast method for paper currency number recognition on RMB is introduced by details, including image positioning, image segmentation and number recognition. On the foundation of verification in PC, it is implemented by transplanting it to DSP processing platform.The algorithm first quickly navigates to the area of the number using hierarchical positioning followed by rotation correction. Then a new mixed method to get single binarized character is proposed, which ensures the performance of recognition in the system. In order to improve the speed of character recognition, the character recognition algorithm used is based on structural features. For similar characters, classifier is designed based on weighted template. The amount of matching calculation is reduced because of tree classifier used before. The experiments show that the algorithm can reduce the time consumption under the premise of ensuring the correct rate. So the method presented in this dissertation has certain advantages in both speed and accuracy rate.