【作者】 付晓全；

【导师】 黄??；

【作者基本信息】 哈尔滨工业大学 ， 计算机科学与技术， 2006， 硕士

【摘要】 纸币图像识别是近年来在模式识别领域较为活跃的一个课题,而且有着很广阔的应用前景,由此技术研制的纸币清分机正在银行等金融系统发挥着越来越大的作用。在进行纸币图像识别时,由于采集到的图像是在低分辨率下扫描得到的,如何在这种低分辨率条件下实现对纸币进行识别,并且识别速度要满足清分机的要求,是纸币识别的核心问题。本文通过对纸币特征进行分析,对纸币的识别过程做了详细研究。在图像的预处理中,由于系统对纸币处理实时性的要求,该过程必须在很短时间内完成,因此我们对图像的边界检测,撇弃了诸如图像二值化,Hough变换等处理手段,而是直接采用分段线性拟合的方法确定图像中纸币的四个边界,以完成纸币的定位和倾斜矫正。在特征提取中,我们对基于方向块的特征提取方法进行了分析,在此基础上针对美元特点,对图像方向块的划分方式做了研究,并提出了基于几何距离的特征提取方法;在分类器设计中,我们采用了LVQ网络对纸币进行学习与分类,并提出了一种具有两层结构的分类算法,第一层首先对输入的特征向量进行粗分类,选定与特征向量匹配距离最大的两类币种,进入第二层分类器;在第二层分类器中,我们通过研究进入该模块两类币种特征块的相关性,重新设计了特征向量,同时对分类器进行改进,最终实现对纸币的分类。在纸币的污损检测方面,提出了一种基于均匀性特征的纸币污损检测方法,首先利用均匀性特征判定待检纸币上可能存在污损的区域,然后利用标准图像进行图像配准,确定这些区域在参考图像上的对应位置,最后逐像素进行比较,最终判定待检图像的污损状况。更多还原

【Abstract】 Paper currency recognition is a popular issue in recent years, because it has a great future in the market. The paper currency sorter developed by designer plays an important role in bank area. Due to the reason of low resolution of paper currency image and high speed of paper currency sorter, so how to recognize paper currency under these conditions is a key problem. Then 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 localization and adjustment directly. By the analysis of based directional block feature extraction, we propose a new method of feature extraction based on geometry distance. In the design of classifier, we use the LVQ network to classify the feature and introduce a classifier with two layers. First classifier chooses two classes whose matching distance between it and paper currency is bigger than others from all class. Then in the second stage, we extract some new feature and improve the classifier to generate the last result. In the stage of defect detection for paper currency, we advances a homogeneity based algorithm for the detection of scratch and cracks appearing on paper currency, in which the homogeneity feature of the sensed paper currency image is first constructed to locate the pixels that probably been polluted, the image registration algorithm is subsequently used to overlay the sensed and reference paper currency image. At last, each probably polluted pixel on the sensed image is compared with its corresponding pixel on the reference image to estimate the contamination level更多还原