脉冲耦合神经网络与小数幂指数滤波器在数字识别中的应用研究
|
摘要
数字识别在模拟人工智能、计算机文字处理等方面具有巨大的应用前景,世界各国的模式识别研究者都为此做出大量的研究,提出了很多图像预处理算法和识别算法。然而无论是哪一种算法都无法做到与人类自身的识别能力相媲美。这激励着模式识别工作者不断地完善目前的图像预处理算法和识别算法。
人眼对观察图像目标的识别主要是通过目标特征信息的提取,而作为图像识别系统尤其是在图像平移、缩放和旋转等畸变情况下,怎样保证识别目标的不变性是我们研究的主要内容,尤其是目标形状特征的提取,是能否良好区分不同目标物体的重要标志。
本文首先对数字识别技术进行了概述,并对本文所用的基础知识:脉冲耦合神经网络PCNN以及小数幂指数滤波器FPF的基本概念进行了介绍,然后在大量的前人工作基础上,本文提出了基于PCNN与FPF相结合的一种图像识别算法。其主要思想是将PCNN提取图像边缘信息的特点和FPF对于图像的识别能力进行了有机的结合,根据不同识别目标,适当调整滤波器参数,来达到最佳识别效果。最后,通过MATLAB平台仿真实验表明:相比于经典的MACE滤波器,本算法功能实现效率高,峰值输出明显,判别率有了较大提高。
Numerals Recognition has a prosperous future in the field of simulating artificial intelligence and computer words processing.Researchers worldwide worked on pattern recognition and proposed many algorithms of image preprocessing and pattern recognition.But none can compare with the recognition ability of human beings.This inspires researchers to improve the image pre-processing algorithm,feature extraction algorithm and recognition algorithm.
The image recognition for mankind is mainly dependent on the feature extraction of information.The main problem in target recognition is that images of targets will change with translation,rotation scale and intensity,etc and feature extraction of information is one of the most important steps to solve.But relatively few researches has been done that uses PCNN in numerals recognition.
The main problem in target recognition is that images of targets will change with translation,rotation scale and intensity.A numerals recognition model based on PCNN(Pulse-Coupled Neural Network) and FPF (Fractional-Power Filter) is proposed.This method uses inherent ability of PCNN to extract feature and capability of FPF,then FPF allows invariance to be built into and it can recognize numerals with distortion effectively.The results of computer simulation show that the proposed method has a better effect compared with classical filters such as MACE. The simulation results of 340 images of the numerals from 0 to 9 with translation,rotation and scale demonstrate that the method works well and gets high distinguishing rate.
人眼对观察图像目标的识别主要是通过目标特征信息的提取,而作为图像识别系统尤其是在图像平移、缩放和旋转等畸变情况下,怎样保证识别目标的不变性是我们研究的主要内容,尤其是目标形状特征的提取,是能否良好区分不同目标物体的重要标志。
本文首先对数字识别技术进行了概述,并对本文所用的基础知识:脉冲耦合神经网络PCNN以及小数幂指数滤波器FPF的基本概念进行了介绍,然后在大量的前人工作基础上,本文提出了基于PCNN与FPF相结合的一种图像识别算法。其主要思想是将PCNN提取图像边缘信息的特点和FPF对于图像的识别能力进行了有机的结合,根据不同识别目标,适当调整滤波器参数,来达到最佳识别效果。最后,通过MATLAB平台仿真实验表明:相比于经典的MACE滤波器,本算法功能实现效率高,峰值输出明显,判别率有了较大提高。
Numerals Recognition has a prosperous future in the field of simulating artificial intelligence and computer words processing.Researchers worldwide worked on pattern recognition and proposed many algorithms of image preprocessing and pattern recognition.But none can compare with the recognition ability of human beings.This inspires researchers to improve the image pre-processing algorithm,feature extraction algorithm and recognition algorithm.
The image recognition for mankind is mainly dependent on the feature extraction of information.The main problem in target recognition is that images of targets will change with translation,rotation scale and intensity,etc and feature extraction of information is one of the most important steps to solve.But relatively few researches has been done that uses PCNN in numerals recognition.
The main problem in target recognition is that images of targets will change with translation,rotation scale and intensity.A numerals recognition model based on PCNN(Pulse-Coupled Neural Network) and FPF (Fractional-Power Filter) is proposed.This method uses inherent ability of PCNN to extract feature and capability of FPF,then FPF allows invariance to be built into and it can recognize numerals with distortion effectively.The results of computer simulation show that the proposed method has a better effect compared with classical filters such as MACE. The simulation results of 340 images of the numerals from 0 to 9 with translation,rotation and scale demonstrate that the method works well and gets high distinguishing rate.
引文
[1]Nagy G,Spatial Sampling Effects in Optical Character Recognition[J],ICDAR95,pp.309.
[2]Ibarra O.M,Handwritten digit recognition by means of holographic associative memory.Expert Systems with Applications.1998.15(3):399-403.
[3]官淑兰,手写数字识别的应用与研究,山东大学硕士学位论文,2006.8.
[4]盛寁,刘伟.计算机文字识别的发展及应用[J].科技信息,2008年.31:65.
[5]童继进,刘忠.基于Zernike矩、粗集和神经网络的数字识别方法[J].计算机信息,2005,21(12):172-174.
[6]邬建瓴,数字识别的应用,华中科技大学硕士论文,2006.5.
[7]王永乾,吕蓉。基于BP网络的手写体数字识别方法[J].山东点子,2004,3:24-26.
[8]黄心哗,王茂祥,富煜清.基于结构分析的手写数字识别算法[J]。电子工程师,1999.11:23-25.
[9]高彤,姜华,吕民。基于模板匹配的手写体字符识别方法[J]。哈尔滨工业大学学报,1999,31(1):104-106.
[10]刘建华,牛秦洲,程小辉等.基于特征的印刷体数字符号识别系统[J].桂林工学院学报,2005,25(1):1-3.
[11]马义德,徐光柱,齐春亮,等.PCNN与传统神经网络在图像处理中的应用研究[J].中国科技论文在线,2005,3.
[12]Johnson JL,Ranganath H,Kuntimad G,et al.Pulse Coupled Neural Networks[J].Neural Networks and Pattern Recognition.San Diego,CA:Academicpp,1998:51-56.
[13]Eckhorn R,Reitboeck HJ,Arndt M.Feature linking via synchronization among distributed assemblies:Simulation of results from cat cortex[J].Neural Compute,1990,2(3):293-307.
[14]Johnson J L.Pulse Coupled Neural Net:Translation,rotation,scale,distortion and Intensity signal invariance for images[J].Appl.Opt.,1994,33(26):6239-6253.
[15]Ranganath H S,Kuntimad G,and Johnson J L.Pulse Coupled Neural Networks for image processing[C].In:Proc.1995 IEEE Southeast Con.,Raleigh NC,1995:37-43.
[16]John J L,Ritter D.Observation of periodic waves in a Pulse-Coupled Neural Network[J].Opt.Lett.,1993,18(15):1253-1255.
[17]Kuntimad G,Ranganath H S.Perfect image segmentation using Pulse Coupled Neural Networks[J].IEEE Transactions on Neural Networks,1999,10(3):591-598.
[18]林冬梅.PCNN与粗集理论在生物图像处理中的应用研究[D].兰州:兰州大学硕士学位论文,2009
[19]齐春亮.脉冲耦合神经网络模型参数自动标定算法研究[D].兰州:兰州大学硕士学位论文,2006.
[20]史飞.脉冲耦合神经网络在图像处理中的应用研究[D].兰州:兰州大学硕士学位论文,2003.
[21]Ma Y,Shi F,Li L.A new kind of impulse noise filter based on PCNN[C].IEEE International Conference on Neural Network and Signal Processing,2003:152-155.
[22]Ma Yi-de,Shi Fei,Li Lian.Gaussian noise filter based on PCNN[C].IEEE International Conference on Neural Network and Signal Processing,2003:149-151.
[23]Zhang J,Lu Z,Shi L,et al.Filtering images contaminated with Pep and Salt type noise with Pulse-Coupled Neural Networks[J].Science In China Series F-Information Sciences,2005,48(3):322-334.
[24]Takashi Watanabe,Masaru Tanaka,Takio Kurita,Taketoshi Mishima.Autonomous foveating system based on Pulse-Coupled Neural Network[C].Proceedings of the ITC-CSCC99,1999,01:197-200.
[25]Waldemark J,Becanovic V,Lindblad T,et al.Hybrid neural networks automatic target recognition[C].In IEEE 1997 international conference on system,man,andcybernetics Hyatt Orlando,Florida,USA,1997,10:12-15.
[26]Anuradha G,Priscilla Calvin,G Yuen,et al.Pulse Coupled Neural Network based image classification[C].Proceedings of the Thirtieth Southeastern Symposium on System theory,1998:402-406.
[27]Caulfield H J,Kinser J M.Finding The shortest path in the shortest time using PCNN's[J].IEEE Transactions on Neural Networks,1999,10(3):604-606.
[28]Takashi Watanable,Masaru Tanaka,Takio Kurita,Taketoshi Mishima.Autonomous foveating system based on Pulse-Coupled Neural Network[C].Proceedings of the ITC-CSCC99,1999,01:197-200.
[29]Broussard R P,Rogers S K,Oxley M E et al.Physiologically motivated image fusion for object detection using a Pulse Coupled Neural Network[J].IEEE Trans.Neural Networks,1999,10(3):554-563.
[30]LI W,Zhu X,A new image fusion algorithm based on wavelet packet analysis and PCNN[C].Proceeding of the Fourth IEEE International Conference on Machine Learning and Cybernetics,Guangzhou,China,2005,9:5297-5301.
[31]冈萨雷斯 等著;阮秋琦等译.数字图像处理(第二版)北京:电子工业出版社,2003.3.
[32]Kumar B.V.K.V,Tutorial Survey of Composite Filter Designs for Optical Correlators,Neural Networks,IEEE Trans.Appl.Opt.31(23) 4773-4801(1992).
[33]Lindblad T,Kinser M,Image Processing Using Pulse-Coupled Neural Networks (Second Revised Edition) Springer,2005.9.
[34]Brasher J and Kinser J.M,Fractional-Power Synthetic Discrimininant Functions,Pattern Recognition 27(4),577-585(1994).
[35]张其科 基于遗传神经网络的手写体数字识别研究,中国地质大学硕士毕业论文,2007.5.
[36]Kinser J.M.Foveation by a Pulse-Coupled Neural Network.Neural Networks,IEEE Trans.on(10)3,May 1999 Page(s):621-625.
[37]Christophe Auzizeau.Pulse Coupled Neural Networks for Generation of Video Object Plane.The University of Strathclyde Supervisor:Dr J.Soraghan &Prof.O.Destorages,1999
[2]Ibarra O.M,Handwritten digit recognition by means of holographic associative memory.Expert Systems with Applications.1998.15(3):399-403.
[3]官淑兰,手写数字识别的应用与研究,山东大学硕士学位论文,2006.8.
[4]盛寁,刘伟.计算机文字识别的发展及应用[J].科技信息,2008年.31:65.
[5]童继进,刘忠.基于Zernike矩、粗集和神经网络的数字识别方法[J].计算机信息,2005,21(12):172-174.
[6]邬建瓴,数字识别的应用,华中科技大学硕士论文,2006.5.
[7]王永乾,吕蓉。基于BP网络的手写体数字识别方法[J].山东点子,2004,3:24-26.
[8]黄心哗,王茂祥,富煜清.基于结构分析的手写数字识别算法[J]。电子工程师,1999.11:23-25.
[9]高彤,姜华,吕民。基于模板匹配的手写体字符识别方法[J]。哈尔滨工业大学学报,1999,31(1):104-106.
[10]刘建华,牛秦洲,程小辉等.基于特征的印刷体数字符号识别系统[J].桂林工学院学报,2005,25(1):1-3.
[11]马义德,徐光柱,齐春亮,等.PCNN与传统神经网络在图像处理中的应用研究[J].中国科技论文在线,2005,3.
[12]Johnson JL,Ranganath H,Kuntimad G,et al.Pulse Coupled Neural Networks[J].Neural Networks and Pattern Recognition.San Diego,CA:Academicpp,1998:51-56.
[13]Eckhorn R,Reitboeck HJ,Arndt M.Feature linking via synchronization among distributed assemblies:Simulation of results from cat cortex[J].Neural Compute,1990,2(3):293-307.
[14]Johnson J L.Pulse Coupled Neural Net:Translation,rotation,scale,distortion and Intensity signal invariance for images[J].Appl.Opt.,1994,33(26):6239-6253.
[15]Ranganath H S,Kuntimad G,and Johnson J L.Pulse Coupled Neural Networks for image processing[C].In:Proc.1995 IEEE Southeast Con.,Raleigh NC,1995:37-43.
[16]John J L,Ritter D.Observation of periodic waves in a Pulse-Coupled Neural Network[J].Opt.Lett.,1993,18(15):1253-1255.
[17]Kuntimad G,Ranganath H S.Perfect image segmentation using Pulse Coupled Neural Networks[J].IEEE Transactions on Neural Networks,1999,10(3):591-598.
[18]林冬梅.PCNN与粗集理论在生物图像处理中的应用研究[D].兰州:兰州大学硕士学位论文,2009
[19]齐春亮.脉冲耦合神经网络模型参数自动标定算法研究[D].兰州:兰州大学硕士学位论文,2006.
[20]史飞.脉冲耦合神经网络在图像处理中的应用研究[D].兰州:兰州大学硕士学位论文,2003.
[21]Ma Y,Shi F,Li L.A new kind of impulse noise filter based on PCNN[C].IEEE International Conference on Neural Network and Signal Processing,2003:152-155.
[22]Ma Yi-de,Shi Fei,Li Lian.Gaussian noise filter based on PCNN[C].IEEE International Conference on Neural Network and Signal Processing,2003:149-151.
[23]Zhang J,Lu Z,Shi L,et al.Filtering images contaminated with Pep and Salt type noise with Pulse-Coupled Neural Networks[J].Science In China Series F-Information Sciences,2005,48(3):322-334.
[24]Takashi Watanabe,Masaru Tanaka,Takio Kurita,Taketoshi Mishima.Autonomous foveating system based on Pulse-Coupled Neural Network[C].Proceedings of the ITC-CSCC99,1999,01:197-200.
[25]Waldemark J,Becanovic V,Lindblad T,et al.Hybrid neural networks automatic target recognition[C].In IEEE 1997 international conference on system,man,andcybernetics Hyatt Orlando,Florida,USA,1997,10:12-15.
[26]Anuradha G,Priscilla Calvin,G Yuen,et al.Pulse Coupled Neural Network based image classification[C].Proceedings of the Thirtieth Southeastern Symposium on System theory,1998:402-406.
[27]Caulfield H J,Kinser J M.Finding The shortest path in the shortest time using PCNN's[J].IEEE Transactions on Neural Networks,1999,10(3):604-606.
[28]Takashi Watanable,Masaru Tanaka,Takio Kurita,Taketoshi Mishima.Autonomous foveating system based on Pulse-Coupled Neural Network[C].Proceedings of the ITC-CSCC99,1999,01:197-200.
[29]Broussard R P,Rogers S K,Oxley M E et al.Physiologically motivated image fusion for object detection using a Pulse Coupled Neural Network[J].IEEE Trans.Neural Networks,1999,10(3):554-563.
[30]LI W,Zhu X,A new image fusion algorithm based on wavelet packet analysis and PCNN[C].Proceeding of the Fourth IEEE International Conference on Machine Learning and Cybernetics,Guangzhou,China,2005,9:5297-5301.
[31]冈萨雷斯 等著;阮秋琦等译.数字图像处理(第二版)北京:电子工业出版社,2003.3.
[32]Kumar B.V.K.V,Tutorial Survey of Composite Filter Designs for Optical Correlators,Neural Networks,IEEE Trans.Appl.Opt.31(23) 4773-4801(1992).
[33]Lindblad T,Kinser M,Image Processing Using Pulse-Coupled Neural Networks (Second Revised Edition) Springer,2005.9.
[34]Brasher J and Kinser J.M,Fractional-Power Synthetic Discrimininant Functions,Pattern Recognition 27(4),577-585(1994).
[35]张其科 基于遗传神经网络的手写体数字识别研究,中国地质大学硕士毕业论文,2007.5.
[36]Kinser J.M.Foveation by a Pulse-Coupled Neural Network.Neural Networks,IEEE Trans.on(10)3,May 1999 Page(s):621-625.
[37]Christophe Auzizeau.Pulse Coupled Neural Networks for Generation of Video Object Plane.The University of Strathclyde Supervisor:Dr J.Soraghan &Prof.O.Destorages,1999
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |