Research On Serial Number Identification Of RMB Banknote Based On Support Vector Machines

As the computer technology development, character recognition technology has been gradually move to maturity, especially for the character recognition, has reached high recognition. Paper currency identification, a scared samples, nonlinear and high dimensions pattern recognition problem is an important part of discriminating system and one of the difficult problems of modern pattern recognition.RMB image pre-processing and serial number identification are achieved. This paper mainly studied the sequential minimal optimization algorithm of support vector machine in statistical learning theory. The support vector machine composed by least sequence algorithm has been used in serial number identification. The paper mainly consists of the following parts:1,Raised the necessity of this subject for research on the basis of the extensive application in identification system researching and the current domestic situation of the image Introduce the image of the basic characteristics, pattern recognition of the basic theory and the issue of identification of the system architecture. This article mainly introduced the basic characteristics of RMB image, the basic theory of pattern recognition and the system architecture of the research.2,The main content of Statistical Learning Theory and the idea of support vector machine algorithm are introduced in the article. Including the key content of Statistical Learning Theory, derived the case of support vector machine algorithm from linear separable to nonseparable, The key concept and characteristics of support vector machine algorithm is descript in the article. Discussed some kinds of support vector machine algorithm in detail, analyzed quadratic programming algorithm and compared analysis the classification performance of several kinds about decomposition method. And then introduced support vector machine algorithm of this subject.3,Introduced the sequence number of raw data sampling, pretreating and feature extraction method. And then, support vector machine composed by least sequence algorithm has been used in serial number identification. Simulated result using Matlab and the experimental result is also on the paper. The experimental result indicates that this recognition method has a higher monetary recognition and identification than simple BPNN.