Research Paper Currency Number Recognition Under Complex Background

Crown word numbers on each bill is unique, so you can use it to identify the notes. In practical applications, Identification process must meet the requirements of accuracy and real-time. According to the actual situation, Study a fast recognition algorithm of complex background of bank note number, and apply it to the paper money identification device, that have very strong practicability and significance of the research. This subject is to complete the system design from the paper currency number image acquisition to character number recognition, including: image acquisition, image graying, gray histogram adjustment, image enhancement, image positioning, image segmentation, character recognition.Firstly, this paper chooses DH-VT120 image acquisition card Daheng Group to gain bill samples images. In order to reduce the amount of calculation and computation time, we need to achieve gray scale using the weighted method, and then realize the number of regional coarse positioning of the gray image.Secondly, the image gray histogram adjustment are studied. This paper introduces three kinds of methods: linear method, nonlinear method, average method. Through the experimental contrast effect. Method of this thesis is the logarithm of the nonlinear adjustment. The method makes the low gray value of the image is expanded, high gray value are compressed.Again, In image enhancement, Using the method of multi-scale wavelet analysis and gray morphological reconstruction respectively experiments. Through the data and processing results, we found that the grayscale morphological reconstruction method is more suitable for this system. This method can not only highlight the target area, but also can blur the background region, for the broken stroke also have good remedial effect.Finally, this paper select the projection method when you want to achieve number region of fine location and character segmentation.this article has carried on the related research on banknotes numbers recognition algorithm. This paper uses the method of template method rough matching and fine character recognition. The method greatly improves the recognition rate of characters similar structure, furthermore for the recognition effect of digital characters and alphabetic character mixed is very good. Can meet the requirements of the experiment.The algorithms are transplanted to the banknote sorter system using DM642 chip, through debugging and verification system. The results show that: number recognition algorithm proposed can meet the recognition sorter real-time and accurate requirements..