面向嵌入式应用的货币智能鉴别关键技术研究

Paper currency identification techniques, which conclude banknote's authenticity, denomination, old or new, impairment, orientation and series, are the common key techniques applied in financial self-service and paper currency recognition devices. There are three main problems in embedded paper currency identification such as low rate of identification to receive, high rate of error resolution, and low recognition speed. So, this project will perform research from aspects listed below. First, by trying to introduce tactile computing technology to detect characteristics of paper currency,it will study new currency counterfeiting detection technology to further expand the currency dimension feature information.Second, it will study damaged banknote's image degradation model and recovery method so as to increase self-adaptiveness in the identification process. These two research results may be helpful to increase the receive rate and reduce the error rate. Third, aiming at the embedded application background, by studying the currency holographic feature information extraction and efficient representation method, it will build an organic, open, long-life-cycle currency holographic features data model. Based on this model, it will study the parallel optimal identification algorithm to improve the identification rate. The research results of this project are expected to provide theoretical and methodological guidance for rapid and reliable paper currency identification applied in the resource constraints of embedded systems.

货币鉴别技术是各类金融自助服务以及银行业务货币鉴别设备的共性关键技术，其内容包括纸币的真伪、面额、新旧、残缺、朝向、版本识别等。目前国内外嵌入式货币鉴别应用普遍存在着鉴别接收率较低、误辨率较高、鉴别速度较慢等三大主要问题。为此，本课题研究货币防伪检测技术新方法，通过引入计算触觉技术来检测货币纸质特征，进一步拓展货币特征信息维度；同时，研究残损纸币的图像退化模型及恢复方法，从而提高在鉴别过程中的自适应性；以上两方面的研究成果将有望提高鉴别接收率，并降低误辨率。此外，针对嵌入式应用背景，将研究完善货币全息特征信息提取与高效表示方法，构建一个有机、开放、长生命周期的货币全息特征数据模型；并基于该全息特征数据模型，研究典型嵌入式应用背景下的并行最优鉴别算法，提高鉴别速度。本项目的研究成果将有望为嵌入式应用背景下货币的快速、可靠鉴别提供理论指导和方法指导。

项目构建了一个基于光纤传感的纸币触觉检测装置，一定程度可检测纸币纹理粗细粒度，拓展机器检测特征信息的维度；项目构建了残损纸币尤其是折叠纸币的图像退化模型与恢复算法，可显著提高真币接受率；基于可见光、红外、紫外、磁性等纸币特征信息的提取，构建了人民币、美元等某些面值纸币的特征数学模型，以此为基础，针对一类嵌入式货币识别系统进行了相关鉴别实验研究，表明了相关模型与方法在识别速度与鉴别率等方面的可行性。