纸币数字的多特征综合识别算法-计算机工程与应用.PDFVIP

Computer Engineering and Applications 计算机工程与应用 2014 ，50（19） 169 纸币数字的多特征综合识别算法 杜 鑫，刘文波，李开宇 DU Xin, LIU Wenbo, LI Kaiyu 南京航空航天大学 自动化学院，南京 210016 College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China DU Xin, LIU Wenbo, LI Kaiyu. Complex identification algorithm with multi-feature via banknotes number. Com- puter Engineering and Applications, 2014, 50 （19）：169-172. Abstract ：A novel complex identification approach with multi-feature based on banknotes number is designed, which addresses the slow rate and low efficiency of paper currency recognition. Firstly, topology of banknote character image is analyzed. The structural features, such as endpoint of the character pixel, have been processed. Then, the character topo- logical features have been modified, and its results have been more classified and distinguished. In addition, as the damaged character does not satisfy defaced character of the topology, template matching method has been introduced to identify. Finally, the experimental results show that compared with traditional identification algorithm, the proposed method has better efficiency and faster rate. Key words ：banknote; digits recognition; topology; template matching 摘 要：针对目前存在的纸币识别率低和速度慢的问题，提出了一种新型的纸币数字多特征综合识别算法。分析细 化后纸币字符图像的拓扑结构，以及字符像素的端点等结构特征，对其进行初步分类；改变字符拓扑特征，对上一步 骤的结果进行进一步分类识别；针对破损字符不满足拓扑结构特征的污损字符，采用了模板匹配的方式进行了补漏 识别。实验结果表明，相对于传统的识别算法，提出的算法不仅具有更高的识别率，而且具有较高的识别速率。 关键词：纸币；数字识别；拓扑结构；模板匹配 文献标志码：A 中图分类号：TP301 doi ：10.3778/j.issn. 1002-8331.1211-0033 1 引言 写体数字有较好的识别率。文献[7]提出了一种基于神 人民币编码是同一批发行的人民币的编号，也被称 经网络的纸币识别算法，该算法的优点在于识别率高， 为冠号。冠号用于区别纸币的印刷批次、范围、计划，以 对图像污损抗干扰能力强。目前主流的纸币清分器均 及印刷车间甚至印刷厂。同时，冠号在某些方面，如标 采用该方法。但由于神经网络训练样本周期长，对于没 识纸币的身份，电子识别，对人民币的防伪和收藏等方 有训练过的图识别效率不高，所以该方法仍然需要进一 面具有着重要的价值。