纸币清分机图象识别系统的研究与设计
|
摘要
图像识别是目前具有广阔市场应用前景和发展空间的识别技术之一,而广泛应用在金融领域中的纸币清分机对纸币图象识别与处理技术有着更为严格的要求。纸币清分机的的核心技术就是必须具有实时的纸币图象检测硬件电路和高效的纸币图象识别算法。本文针对纸币图象检测和处理问题进行了深入研究,提出并设计了一种具有多功能的纸币清分机的图象检测硬件电路和相应的图象识别方法,满足了纸币清分的实时性和准确性的技术要求。
在检测系统硬件电路的设计上,本文提出了将接触式图像传感器(CIS)、复杂可编程逻辑器件(CPLD)与数字信号处理(DSP)技术相结合的设计方案:纸币首先通过CIS进行高速图象采集和A/D转换电路的转换,将纸币图象数据存入DSP的存储空间;充分利用DSP的高速信号处理功能,选择TI公司的TMS320C6711芯片完成图像数据的必要处理,进而实现纸币的识别与清分功能;利用CPLD的高速逻辑信号处理能力,选择Xilinx公司的XC95144完成系统硬件的时序控制并提供电路所需的逻辑信号。
在纸币图象识别算法上,以国内目前流通的第四版和第五版人民币作为识别对象,应用数字图像处理技术与模式识别方法,并结合改进的自组织映射(SOFM)神经网络,来实现人民币的实时识别,最终实现的识别算法具有较强的实时性。该算法以人民币图像的灰度特征信号为基础,完成纸币图像定位和倾斜校正、纸币的尺寸测量、纸币图像的特征值提取等一系列的检测任务,并根据检测到的数据识别纸币的面值、版本、正反面、正倒向、缺损、脏污等,进而对纸币加以识别及清分。
实验结果表明,本文提出的纸币清分机方案和所应用的纸币图象检测硬件电路及纸币图象识别算法满足纸币识别及清分的技术要求,具有较高的实用价值。
Paper currency recognition is a kind of technology about image processing which is widely used at present. It has a wide market prospect and development space especially in bank. The bill sortors which are used in optical financial facility widely provide most strick demand about the bill image recognition and processing technolugy. The key point of studying the bill sortor system is the design of hardware circuit and high efficient recognition algorithm. This thesis makes a deep research on these problems. Finally, a kind of hardware circuit and its corresponding recognition method for paper currency are provided in this thesis. The system can meet the technology level of real time and accuracy.
In the design of test system hardware circuit, the technologies of CIS, complex programmable logic device (CPLD) and digital signal processor (DSP) are combined. After high speed collection through CIS and conversion with A/D, the bill image datas are saved in the DSP space, selected TI compony TMS320C6711 as DSP, which used high speed digital signal processing function fully, and the DSP gives the image datas a series of disposal in order to realize the recognition and separation. Selected Xilinx compony XC95144 as CPLD, which used high speed logic signal processing function fully, and the CPLD provides the clock and logical signals for the whole system.
In the bill image recognition arithmetic design, digital image processing technique, pattern recognition and improved SOFM neural network are used to recognize the current 4th and 5th RMB images. The recognition method has a strong adaptability. It is based on the gray level feature of RMB images. After the steps of orientation, lean adjustment, measure of the size and extraction of the image feature, we can recognize its value, version, direction, deformity and dirty, etc. Furthmore, bill are recognized and sorted.
As a result, the experiment proves that the bill sortor project system, its bill image test hardware circuit andbill image recognition arithmetic can meet the technology requirements, and it has high practical values.
在检测系统硬件电路的设计上,本文提出了将接触式图像传感器(CIS)、复杂可编程逻辑器件(CPLD)与数字信号处理(DSP)技术相结合的设计方案:纸币首先通过CIS进行高速图象采集和A/D转换电路的转换,将纸币图象数据存入DSP的存储空间;充分利用DSP的高速信号处理功能,选择TI公司的TMS320C6711芯片完成图像数据的必要处理,进而实现纸币的识别与清分功能;利用CPLD的高速逻辑信号处理能力,选择Xilinx公司的XC95144完成系统硬件的时序控制并提供电路所需的逻辑信号。
在纸币图象识别算法上,以国内目前流通的第四版和第五版人民币作为识别对象,应用数字图像处理技术与模式识别方法,并结合改进的自组织映射(SOFM)神经网络,来实现人民币的实时识别,最终实现的识别算法具有较强的实时性。该算法以人民币图像的灰度特征信号为基础,完成纸币图像定位和倾斜校正、纸币的尺寸测量、纸币图像的特征值提取等一系列的检测任务,并根据检测到的数据识别纸币的面值、版本、正反面、正倒向、缺损、脏污等,进而对纸币加以识别及清分。
实验结果表明,本文提出的纸币清分机方案和所应用的纸币图象检测硬件电路及纸币图象识别算法满足纸币识别及清分的技术要求,具有较高的实用价值。
Paper currency recognition is a kind of technology about image processing which is widely used at present. It has a wide market prospect and development space especially in bank. The bill sortors which are used in optical financial facility widely provide most strick demand about the bill image recognition and processing technolugy. The key point of studying the bill sortor system is the design of hardware circuit and high efficient recognition algorithm. This thesis makes a deep research on these problems. Finally, a kind of hardware circuit and its corresponding recognition method for paper currency are provided in this thesis. The system can meet the technology level of real time and accuracy.
In the design of test system hardware circuit, the technologies of CIS, complex programmable logic device (CPLD) and digital signal processor (DSP) are combined. After high speed collection through CIS and conversion with A/D, the bill image datas are saved in the DSP space, selected TI compony TMS320C6711 as DSP, which used high speed digital signal processing function fully, and the DSP gives the image datas a series of disposal in order to realize the recognition and separation. Selected Xilinx compony XC95144 as CPLD, which used high speed logic signal processing function fully, and the CPLD provides the clock and logical signals for the whole system.
In the bill image recognition arithmetic design, digital image processing technique, pattern recognition and improved SOFM neural network are used to recognize the current 4th and 5th RMB images. The recognition method has a strong adaptability. It is based on the gray level feature of RMB images. After the steps of orientation, lean adjustment, measure of the size and extraction of the image feature, we can recognize its value, version, direction, deformity and dirty, etc. Furthmore, bill are recognized and sorted.
As a result, the experiment proves that the bill sortor project system, its bill image test hardware circuit andbill image recognition arithmetic can meet the technology requirements, and it has high practical values.
引文
铩颷1]和均,王凤英.银行现金清分机现状及需求的分析与调查.金融&科技,2004年1月,第1期.
[2]哈尔滨彼绍特科技开发有限公司市场营销部.松花江CF2000型纸币清分机资料汇编,2003年8月,第四版.
[3]娄小冬.智能纸币清分机系统设计及其实现策略研究.杭州:浙江大学,2005年.
[4]陈慧鹏.智能型纸币清分机图像识别处理方法研究.武汉:华中科技大学,2005年.
[5]欧阳松林.“金通”现金清分机的研制.电脑与信息技术,2000年,第2期:45-47.
[6]欧阳松林.多分类器融合在银行票据识别中的应用.计算机工程与应用,2002年8月.
[7]王庆文.关于中小城市票据自动清分系统业务需求分析.现代电子技术,2003年第2期:79-82.
[8]王德福,梁士文.铁路客票自动售票机与现金识别处理技术.计算机工程与应用,1999年7月:12-14.
[9]刘家锋,刘松波,唐降龙.一种实时纸币识别方式的研究.计算机研究与发展,2003年7月:1057-1061.
[10]Fumiaki Takeda,Sigeru Omatu.High speed paper currency recognitionby neural networks.IEEE Yrans on Neural Network,1995,6(1):73-77.
[11]黄德双.神经网络模式识别系统理论.北京:电子工业出版社,1996.
[12]刘贵喜等.CIS信号检测与处理技术.光电工程,2000,27(4);18-20.
[13]周翟和等.纸币号码读入识别系统的单片机实现.工业控制计算机,2002,15(1):69-71.
[14]David J.DeFatta,Joseph G.Lucas,William S.Hodgkiss.Digital SignalProcessing:A system Design Approach.World Pub.Co.,1988.
[15]李方慧等.TMS320C6000系列DSPs原理与应用(第2版).北京:电子工业出版社,2003.
[16]TMS320C6000 CPU and Instruction Set Reference Guide.TexasInstruments Incorporated,January 2001.
铩颷17]潘松,黄继业.EDA技术实用教程.北京:科学出版社,2002.
[18]PI240MC-A4 DataSheet.Peripheral Imaging Corporation,2001.
[19]TMS320C6711 FLOATING-POINT DIGITAL SIGNAL PROCESSOR.Texas Instruments Incorporated,1999.
[20]TMS320C6000 Peripherals Reference Guide.Texas InstrumentsIncorporated,February 2001.
[21]TMS320C6000 Programmer's Guide.Texas Instruments Incorporated,February 2001.
[22]TMS320C6000 Code Composer Studio User's Guide.Texas InstrumentsIncorporated,May 2000.
[23]陈明意.TMS320C6000系列DSP串行通信接口设计.武汉工业学院学报,2004,23(2):7-9.
[24]王宏义,王伟.TMS320C6000系列DSP的自举方式研究.现代电子技术,2003,166(23):36-38.
[25]TMS320C6000 DMA Example.Texas Instruments Incorporated,1999.
[26]XC95144 DataSheet.Xilinx Incorporated,2002.
[27]刘连浩等.智能型点钞机的研制,机电一体化.2003,1:11-13.
[28]夏良正.数字图像处理.南京:东南大学出版社,1999.
[29]Castleman K R.Digital image processing.Electronics Industry Press,1998.
[30]王新成.高级图像处理技术。北京:中国科学技术出版社,2001.
[31]张宏林.Visual C++数字图像模式识别技术及工程实践.北京:人民邮电出版社,2003.
[32]边肇祺,张学工等.模式识别(第二版).北京:清华大学出版社,2000.
[33]李金宗.模式识别导论.北京:高等教育出版社,1994.
[34]Vapnik V.N and Chervonenkis A.Ja.Theory of Pattern Recognition.Nauka,Moscow,1974.
[35]Sonehara.An application of neural network to image data.Trans.Inst.Syst,Control and Info.Eng,1991,35(1):11-18.
[36]Hammerstron D.Working with Neural Network.IEEE Spectrum,1993,30(7):46-53.
[37]J Illing,J Kittler.A survey of the hough transform.Computer on Vision,Graphics,ImageProcessing,1988,44(1):87-116.
铩颷38]A Frosini,M Gori,P Priamil.A neural network-based model for papercurrency recognition and verification.IEEE Trans on Neural Networks,1996,7(6):1482-1490.
[39]赖志国,余啸海.Matlab图像处理与应用.北京:国防工业出版社,2004.
[40]卜飞宇,刘长松,丁晓青.灰度名片图像快速倾斜检测和校正方法.中文信息学报,2004,18(1):62-69.
[41]张跃.一元线性回归的推广.成都教育学院学报,2000,3:69-71.
[42]刘斌昺,曹丹华,吴裕斌.矩形目标的快速恢复方法及其DSP实现.计算机工程,2004,30(1):163-164.
[43]王斌.基于SOFM神经网络的数字模式识别方法.微机发展,2003,13(8):6-7
[44]Kohonen T.The self-Organizing Map.Proc IEEE,1990,78(9):1464-1480.
[45]L.S.Davis,A survey of Edge Detection Techniques,CGIP,4:248-270,1975.2002.
[46]D.MarrandE.Hildreth,Theory of Edge Detection,Proc.R.Soc.London,Ser.B.1980,207:187-217.
[47]Pal N R,Pal S K.A review on emage segmentation techniques.PatternRecognition,1993,26:1277-1294.
[48]Zhang Y J.A survey on evaluation methods for image segmentation.Pattern Recognition,1996,29(8):1335-1346.
[49]Heyden F.Evaluation of edge detection algorithms.Proceedings of3ICIPIA,1990,618-622.
[50]Ng W S,Lee C K.Comment on using the uniformity measure forperformance measure in image segmentation.IEEE-PAMI,1996,18(9):933-934.
[51]Zhang Y J.Influence of segmentation over feature measurement.Pattern Recognition Letters,1995,16:201-206.
[52]Zhang Y J,Luo H T.Optimal selection of segmentation algorithms basedon performance evaluation.Optical Engineering,2000,39(6):1450-1456.
[53]Chalana V,Kim Y.A methodology for evaluation of boundary detection algorithms on medical images. IEEE-MI, 1997, 16(5): 642-652.
[54] TMS320C6000 Programmer's Guide, Texas Instruments, Inc., August 2002.
[55] TMS320C6000 Code Composer Studio Tutorial, Texas Instruments, Inc., February 2000.
[2]哈尔滨彼绍特科技开发有限公司市场营销部.松花江CF2000型纸币清分机资料汇编,2003年8月,第四版.
[3]娄小冬.智能纸币清分机系统设计及其实现策略研究.杭州:浙江大学,2005年.
[4]陈慧鹏.智能型纸币清分机图像识别处理方法研究.武汉:华中科技大学,2005年.
[5]欧阳松林.“金通”现金清分机的研制.电脑与信息技术,2000年,第2期:45-47.
[6]欧阳松林.多分类器融合在银行票据识别中的应用.计算机工程与应用,2002年8月.
[7]王庆文.关于中小城市票据自动清分系统业务需求分析.现代电子技术,2003年第2期:79-82.
[8]王德福,梁士文.铁路客票自动售票机与现金识别处理技术.计算机工程与应用,1999年7月:12-14.
[9]刘家锋,刘松波,唐降龙.一种实时纸币识别方式的研究.计算机研究与发展,2003年7月:1057-1061.
[10]Fumiaki Takeda,Sigeru Omatu.High speed paper currency recognitionby neural networks.IEEE Yrans on Neural Network,1995,6(1):73-77.
[11]黄德双.神经网络模式识别系统理论.北京:电子工业出版社,1996.
[12]刘贵喜等.CIS信号检测与处理技术.光电工程,2000,27(4);18-20.
[13]周翟和等.纸币号码读入识别系统的单片机实现.工业控制计算机,2002,15(1):69-71.
[14]David J.DeFatta,Joseph G.Lucas,William S.Hodgkiss.Digital SignalProcessing:A system Design Approach.World Pub.Co.,1988.
[15]李方慧等.TMS320C6000系列DSPs原理与应用(第2版).北京:电子工业出版社,2003.
[16]TMS320C6000 CPU and Instruction Set Reference Guide.TexasInstruments Incorporated,January 2001.
铩颷17]潘松,黄继业.EDA技术实用教程.北京:科学出版社,2002.
[18]PI240MC-A4 DataSheet.Peripheral Imaging Corporation,2001.
[19]TMS320C6711 FLOATING-POINT DIGITAL SIGNAL PROCESSOR.Texas Instruments Incorporated,1999.
[20]TMS320C6000 Peripherals Reference Guide.Texas InstrumentsIncorporated,February 2001.
[21]TMS320C6000 Programmer's Guide.Texas Instruments Incorporated,February 2001.
[22]TMS320C6000 Code Composer Studio User's Guide.Texas InstrumentsIncorporated,May 2000.
[23]陈明意.TMS320C6000系列DSP串行通信接口设计.武汉工业学院学报,2004,23(2):7-9.
[24]王宏义,王伟.TMS320C6000系列DSP的自举方式研究.现代电子技术,2003,166(23):36-38.
[25]TMS320C6000 DMA Example.Texas Instruments Incorporated,1999.
[26]XC95144 DataSheet.Xilinx Incorporated,2002.
[27]刘连浩等.智能型点钞机的研制,机电一体化.2003,1:11-13.
[28]夏良正.数字图像处理.南京:东南大学出版社,1999.
[29]Castleman K R.Digital image processing.Electronics Industry Press,1998.
[30]王新成.高级图像处理技术。北京:中国科学技术出版社,2001.
[31]张宏林.Visual C++数字图像模式识别技术及工程实践.北京:人民邮电出版社,2003.
[32]边肇祺,张学工等.模式识别(第二版).北京:清华大学出版社,2000.
[33]李金宗.模式识别导论.北京:高等教育出版社,1994.
[34]Vapnik V.N and Chervonenkis A.Ja.Theory of Pattern Recognition.Nauka,Moscow,1974.
[35]Sonehara.An application of neural network to image data.Trans.Inst.Syst,Control and Info.Eng,1991,35(1):11-18.
[36]Hammerstron D.Working with Neural Network.IEEE Spectrum,1993,30(7):46-53.
[37]J Illing,J Kittler.A survey of the hough transform.Computer on Vision,Graphics,ImageProcessing,1988,44(1):87-116.
铩颷38]A Frosini,M Gori,P Priamil.A neural network-based model for papercurrency recognition and verification.IEEE Trans on Neural Networks,1996,7(6):1482-1490.
[39]赖志国,余啸海.Matlab图像处理与应用.北京:国防工业出版社,2004.
[40]卜飞宇,刘长松,丁晓青.灰度名片图像快速倾斜检测和校正方法.中文信息学报,2004,18(1):62-69.
[41]张跃.一元线性回归的推广.成都教育学院学报,2000,3:69-71.
[42]刘斌昺,曹丹华,吴裕斌.矩形目标的快速恢复方法及其DSP实现.计算机工程,2004,30(1):163-164.
[43]王斌.基于SOFM神经网络的数字模式识别方法.微机发展,2003,13(8):6-7
[44]Kohonen T.The self-Organizing Map.Proc IEEE,1990,78(9):1464-1480.
[45]L.S.Davis,A survey of Edge Detection Techniques,CGIP,4:248-270,1975.2002.
[46]D.MarrandE.Hildreth,Theory of Edge Detection,Proc.R.Soc.London,Ser.B.1980,207:187-217.
[47]Pal N R,Pal S K.A review on emage segmentation techniques.PatternRecognition,1993,26:1277-1294.
[48]Zhang Y J.A survey on evaluation methods for image segmentation.Pattern Recognition,1996,29(8):1335-1346.
[49]Heyden F.Evaluation of edge detection algorithms.Proceedings of3ICIPIA,1990,618-622.
[50]Ng W S,Lee C K.Comment on using the uniformity measure forperformance measure in image segmentation.IEEE-PAMI,1996,18(9):933-934.
[51]Zhang Y J.Influence of segmentation over feature measurement.Pattern Recognition Letters,1995,16:201-206.
[52]Zhang Y J,Luo H T.Optimal selection of segmentation algorithms basedon performance evaluation.Optical Engineering,2000,39(6):1450-1456.
[53]Chalana V,Kim Y.A methodology for evaluation of boundary detection algorithms on medical images. IEEE-MI, 1997, 16(5): 642-652.
[54] TMS320C6000 Programmer's Guide, Texas Instruments, Inc., August 2002.
[55] TMS320C6000 Code Composer Studio Tutorial, Texas Instruments, Inc., February 2000.