基于多传感器信息处理纸币检测若干关键技术研究

发布时间：2018-08-04 14:26

【摘要】：随着人民币发行量、流通量不断地增长,人民币国际化的深入推进,我国人民币流通领域点验钞机具、清分机、ATM机等金融设备急需提高其自动化、信息化水平,以确保人民币的流通效率和管理水平,有效打击纸币造假犯罪活动,保障人民币职能作用和金融体系的安全。本文分析了我国金融设备行业现有的人民币纸币检测技术存在的不足与缺陷:光磁特征多采用定性分析及低水平定量分析鉴别、多传感器特征鉴别方法相对孤立及信息融合能力较弱、智能算法复杂度偏高及实时性较差等,并针对传感器特征鉴别方法孤立及信息融合能力较弱等问题,从多传感器纸币检测机电验证平台搭建与基于多传感器信息纸币图像处理算法研究等方面展开了基于多传感器信息处理的纸币检测若干关键技术研究。多传感器纸币检测机电验证平台搭建工作包含机电平台搭建与检测控制系统的软硬件设计。通过对机电验证平台的搭建,实现了机电平台传送机构在检测控制系统控制下传送纸币依次经过机电平台各传感器模组的运动控制过程,实现了纸币光学、磁性及图像等纸币防伪特征原始数据的动态采集与上传至PC保存,为实现纸币防伪特征定量分析鉴别的离线仿真算法研究提供数据支撑。检测控制系统包含机电控制与光磁检测系统和基于FPGA与CIS纸币图像采集系统,2个子系统间充分利用其控制接口与通信接口实现了系统间协同工作与多传感器鉴别信息的共享,为实现多传感器信息融合算法实物仿真研究提供平台条件。在基于多传感器信息纸币图像处理算法研究工作中,首先利用加权最小二乘法对倾斜的纸币图像进行倾斜检测并通过错切变换算法校正纸币图像,确保了纸币图像冠字号码等特征区域的正确分割提取。通过与随机Hough变换和普通最小二乘法倾斜检测实验对比,仿真结果表明本文倾斜检测算法具有计算量小、稳定性好等特点,可快速实现纸币图像倾斜校正。在纸币图像倾斜校正基础上,基于机电验证平台可实现多传感器鉴别信息共享条件,利用磁性特征鉴别信息识别纸币面值面向信息,实现纸币图像冠字号码快速分割提取与识别算法的研究。基于多传感器信息纸币图像特征区域分割提取算法有利于移植到FPGA纸币图像采集系统中,以减少典型嵌入式纸币图像处理系统中DSP处理单元处理任务。本文着重从多传感器纸币检测机电验证平台搭建及基于多传感器信息纸币图像处理算法前期仿真研究等方面展开基于多传感器信息处理纸币检测技术的研究,为多传感器信息融合算法研究提供原始数据与实物仿真验证平台,为简化纸币图像处理算法与流程,合理分配嵌入式纸币图像处理系统各处理单元任务以提高各单元协同工作能力及纸币图像处理效率进行了有益的前期研究尝试。
[Abstract]:With the increasing of RMB circulation and circulation, and the deepening of RMB internationalization, the level of automation and information technology is urgently needed to be improved in China's financial equipment, such as banknote checking machines, sorting machines, ATM machines, and so on, in the area of RMB circulation. In order to ensure the efficiency and management level of RMB circulation, effectively crack down on the criminal activities of banknote counterfeiting, guarantee the function of RMB and the security of financial system. This paper analyzes the shortcomings and defects of the existing RMB banknotes detection technology in China's financial equipment industry: the optical and magnetic characteristics are mostly identified by qualitative analysis and low level quantitative analysis. The multi-sensor feature identification method is relatively isolated and the ability of information fusion is weak, the complexity of intelligent algorithm is on the high side and the real time is poor, and so on, and aiming at the problems of the isolation of sensor feature identification method and the weak ability of information fusion, etc. Several key techniques of paper currency detection based on multi-sensor information processing are studied from the aspects of multi-sensor paper currency detection electromechanical verification platform and image processing algorithm based on multi-sensor information. The multi-sensor paper currency detection and electromechanical verification platform includes the hardware and software design of the electromechanical platform and the detection and control system. Through the construction of the electromechanical verification platform, the paper currency transfer mechanism under the control of the detection and control system realizes the motion control process of each sensor module of the electromechanical platform in turn, and realizes the paper currency optics. The original data of magnetic and image anti-counterfeiting features are dynamically collected and uploaded to PC to provide data support for the off-line simulation research of quantitative analysis and identification of banknote anti-counterfeiting features. The detection and control system includes electromechanical control and optomagnetic detection system and paper money image acquisition system based on FPGA and CIS. The two subsystems make full use of their control interface and communication interface to realize the sharing of cooperative work and multi-sensor identification information between the two subsystems. It provides a platform for the simulation of multi-sensor information fusion algorithm. In the research work of paper currency image processing algorithm based on multi-sensor information, firstly, the weighted least square method is used to detect the tilt of the skewed banknote image and the miscut transform algorithm is used to correct the banknote image. It ensures the correct segmentation and extraction of the feature areas of banknote image, such as crowning and number. Compared with the experiments of random Hough transform and ordinary least square method, the simulation results show that the proposed algorithm has the advantages of small computation and good stability, and can quickly realize the skew correction of paper currency images. Based on the skew correction of banknote image and the electromechanical verification platform, the multi-sensor discriminant information sharing condition can be realized, and the face value information of banknotes can be identified by using magnetic feature discriminant information. Research on fast segmentation and recognition algorithm of banknote image. The feature region segmentation and extraction algorithm based on multi-sensor information paper currency image is advantageous to transplant to FPGA paper currency image acquisition system, in order to reduce the processing task of DSP processing unit in typical embedded paper currency image processing system. This paper focuses on the multi-sensor paper currency detection based on the establishment of electromechanical verification platform and based on multi-sensor information paper currency image processing algorithm simulation research based on multi-sensor information processing paper currency detection technology. In order to simplify the algorithm and flow of paper currency image processing, the paper provides a platform of raw data and physical simulation verification for the research of multi-sensor information fusion algorithm. In order to improve the cooperative ability of each unit and the efficiency of paper money image processing, a beneficial preliminary research attempt was made by allocating the tasks of each processing unit of embedded paper currency image processing system.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9

【参考文献】