人民币面值快速识别算法研究

发布时间：2018-06-14 19:57

本文选题：识别技术 + 面值识别　；参考：《辽宁科技大学》2016年硕士论文

【摘要】：随着科技的迅猛发展和社会的不断进步,现如今识别技术正以惊人的速度发展。识别技术是一个包涵了图像识别技术、指纹识别技术、人脸识别技术、自动识别技术等为一体的现代新型技术。识别技术产业具有不可估计的发展前景,其中数字号码识别技术的发展尤为迅速。随着现代金融业电子化的发展,人民币面值快速识别技术依旧在银行电子化业务系统中扮演着重要角色。本文在查阅国内外参考文献的基础上,针对纸币图像的采集、纸币图像的预处理、纸币图像的面值识别、纸币图像的面向识别以及纸币图像的新旧识别设计了人民币面值快速识别系统。经仿真实验证明,基本满足了对纸币快速识别的要求。本文的核心研究内容为对经过采集和转化后的纸币面值进行图像预处理和识别。其中包括去图像去噪处理,中值滤波,图像的倾斜校正等几部分,然后在对国内传统纸币面值识别方法基础上做一些改进,以第四版与第五版人民币为研究对像,采用图像处理技术与模式识别技术相结合的方法完成对纸币面值的识别,提高了面值识别的准确率,并针对纸币面向识别采用了SOFM神经网络识别技术,照比传统的识别技术做出了一定的改进,在纸币新旧识别方面本文将国内常用识别方法HIS方法、BP-LVQ方法、LVQ方法进行比较研究。实验结果验证了人民币面值快速识别算法并针对以往算法有了明显改进,经过分析比较新旧识别三种方法方法在不同背景下具有不同优点,能满足现代金融业对纸币识别的要求,具有一定的价值。
[Abstract]:With the rapid development of science and technology and social progress, recognition technology is developing at an alarming speed. Recognition technology is a new modern technology which includes image identification technology, fingerprint recognition technology, face recognition technology, automatic identification technology and so on. Recognition technology industry has an inestimable development prospect, especially digital number recognition technology. With the development of modern financial industry, RMB face value recognition technology still plays an important role in the electronic banking business system. On the basis of consulting references at home and abroad, this paper aims at the collection of banknote images, the preprocessing of banknote images, and the recognition of the face value of banknote images. A fast recognition system of RMB face value is designed for paper currency image and the new and old recognition of paper currency image. The simulation results show that it can meet the requirement of paper currency recognition. The core of this paper is image preprocessing and recognition of the collected and converted banknotes. This includes image denoising, median filtering, image skew correction, and so on. Then some improvements are made on the basis of the traditional method of recognizing the face value of domestic banknotes. The fourth and fifth editions of RMB are taken as the research objects. The recognition of banknote face value is accomplished by combining image processing technology with pattern recognition technology, and the accuracy of face value recognition is improved, and the SOFM neural network recognition technology is used for banknote face recognition. Compared with the traditional recognition technology, this paper compares his method with BP-LVQ method and LVQ method in the recognition of new and old banknotes. The experimental results verify the fast recognition algorithm of RMB face value and improve obviously the previous algorithms. After analyzing and comparing the new and old recognition methods, the three methods have different advantages in different backgrounds. To meet the requirements of the modern financial industry for paper money recognition, has a certain value.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】