【作者】 张国华；

【导师】 梁中华；

【作者基本信息】 沈阳工业大学 ， 控制理论与控制工程， 2005， 硕士

【摘要】 纸币是现代金融的血液,是当今世界上必不可少的东西。自建国以来我国共发行了五套人民币,目前正在流通的是第四套和第五套。纸币面额识别是所有金融设备一个必不可少的组成模块,本文在广泛查阅大量国内外参考文献关于纸币面额识别的基础上,通过仔细分析和研究纸币清分机中的面额识别模块,针对我国人民币自身的特征,设计了两种识别方法。 本文的核心技术在于图像的预处理和识别。首先,为了保证识别的准确性,需要对图像进行预处理,如去除噪声、图像增强、定位分割、图像倾斜校正等。在倾斜校正的过程中,本文提出了采用直线拟合的方法计算出边框底线所在直线的斜率,从而计算出纸币图像的倾斜度,然后再进行图像旋转。在图像定位分割中采用了灰度投影的方法,从而准确定位出纸币的上下和左右边界,使图像和背景分离开来。在纸币识别过程中,一方面根据纸币图像的特殊性,提出了采用轴对称掩膜的方法来提取人民币特征;另一方面充分利用神经网络的并行分布处理和高的容错性特点,将BP神经网络算法应用于纸币的面额识别。大量实验结果表明,该识别算法不仅能够识别新币和感染轻微噪声的纸币,而且对缺损较少但仍能流通的残币、旧币仍具有很强的鲁棒性。当一张人民币通过该识别网络后,识别的结果包括:(1)人民币的面额;(2)人民币正、反面。 识别算法中除了采用神经网络进行识别外,还尝试了模板匹配的方法。传统的二维模板匹配虽然实现简单,但计算量十分庞大,花费时间太长。针对此情况提出了采用一维灰度投影的模板匹配方法。实验结果表明,后者在保证了匹配准确率略高的前提下,识别速度也明显高于前者。基于BP神经网络的识别方法由于采用了轴对称掩膜技术来提取纸币特征,所以只需考虑纸币的正面正放和反面正放这两种进入状态即可,与一维投影模板匹配识别一种面额需要考虑正面正放、正面倒放、反面正放、反面倒放相比,在识别率基本相同的情况下,当用于新币种的人民币或其它国家的纸币纸币识别时,推广能力明显高于后者。更多还原

【Abstract】 Banknote is nowdays blood of finance. Since China founded, five series of RMB have been issued, and the fourth or the fifth is now circulating. Denomination identifying of banknote is a necessarily composed module on all financial equipments. On the basis of references about denomination identifying of banknote at abroad and home, by carefully analyzing and studying the corresponding module of the banknote detaching-machine, contraposing to RMB self-characteristics, this thesis designs two recognition methods.The key techniques of the thesis are the image pre-processing and recognition. Firstly, in order to ensure accuracy, it is necessary to pre-process original images. For example, eliminate image noise, enhance image, lacate and segment image, adjust declining image etc. In the course of adjusting declining image, this thesis puts forward a method of calculating the slope of the bottom frame line with linefit, consequently the declining angle is obtained, then we can rotate the image based on the above angle, In the process of recognition, one hand the characteristics are extracted with axis-symmetry masks according to the specialty of the RMB image, on the other hand we adequately use the abilities of parallel processing and high fault tolerance for NN, apply BP algorithm to denomination identifying of banknote. Lots of experiments indicate that the recognition method not only recognizes new and slightly polluted banknotes, but also for defected and old but still circulating banknote is strongly robust, too. After RMB passing by the system, we can get two results: (1) the denomination of RMB, (2) face or inverse of RMB.In recognition algorithm we also try the method of template matching besides neural network. Although the traditional method of two-dimensional template matching is simply achieved, the calculating quantities are much larger, it may take very long time. Under this circumstance, the author puts forward the template matching algorithm based on one-dimensional gray projection. The result of experiments indicates that the latter on the premise that the right ratio of matching is higher , the recognition speed is obviously higher than theformer,too. The method that based on neural network only considers two conveyed directions which are head and reverse because it adopts the technique of axis-symmetry masks. Under the circumstances that the recognition rates doesn’t change, it has better ability of popularization than one-dimensional gray projection which need consider four directions such as head upright, head reverse, tail upright, and tail reverse when it is used in recognizing new kinds of RMB and other countries’ banknote.更多还原