一种去噪去污的人民币序列号分割方法

计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (7): 179-183.

一种去噪去污的人民币序列号分割方法

张杰武1，2，张会林1，2，李伦清1，2

1.上海理工大学光电信息与计算机工程学院，上海 200093
2.上海市现代光学系统重点实验室，上海 200093

出版日期:2015-04-01 发布日期:2015-03-31

Segmentation of RMB serial number with removing noise and pollution

ZHANG Jiewu1，2, ZHANG Huilin1，2, LI Lunqing1，2

1.The College of Optical & Electron Information Engineering, University of Shanghai for Science & Technology, Shanghai 200093, China
2.Shanghai Key Laboratory of Modern Optical System, Shanghai 200093, China

Online:2015-04-01 Published:2015-03-31

摘要/Abstract

摘要： 序列号的准确分割是人民币序列号识别的必要前提，然而人民币图像中引入的各类噪声和污染会严重影响其分割效果，为此提出一种能去除污染和噪声的序列号分割方法。在图像预处理时引入高提升滤波，在增强序列号边缘的同时，滤除各类噪声干扰；在字符分割时利用改进型连通区域算法，找到序列号完整的外围轮廓。在各面值人民币图像组成的数据集上验证该方法的可行性，通过与常用分割法准确率的对比显示该方法的优越性。该方法能对污痕、拆痕、字符粘连等污染的图像进行准确分割，比常用的分割方法具有更好的分割性能。

关键词: 人民币, 序列号分割, 高提升滤波, 连通区域法

Abstract: An accurate segmentation result is necessary to RMB serial number identification, however the results are seriously influenced by the noise and pollution on RMB images. For this reason, this paper presents the high-boost filtering in the image pre-filtering to enhance the edge of the serial number, and filters out all kinds of noise at the same time. During the character segmentation, it uses the improved regional connectivity method to find the serial number of the peripheral outline accurately. Experimental results on various denominations of RMB prove that, this method can get an accurate result on such pollution as dirty, folding-traces, character-adhesion etc., has better performance than commonly used segmentation algorithms.

Key words: RMB, serial number segmentation, high-boost filtering, regional connectivity

张杰武1，2，张会林1，2，李伦清1，2. 一种去噪去污的人民币序列号分割方法[J]. 计算机工程与应用, 2015, 51(7): 179-183.

ZHANG Jiewu1，2, ZHANG Huilin1，2, LI Lunqing1，2. Segmentation of RMB serial number with removing noise and pollution[J]. Computer Engineering and Applications, 2015, 51(7): 179-183.

[1]	霍牛义1，2，3，周少武1，张国云1，2，3，郑腾交1. 基于支持向量机的三维鼻子建模方法研究[J]. 计算机工程与应用, 2014, 50(8): 168-172.
[2]	钱晓东，肖强，罗海燕. 基于改进的RBF神经网络的人民币汇率预测研究[J]. 计算机工程与应用, 2010, 46(10): 229-231.