联合变换相关器图像识别的研究
|
摘要
光学相关识别技术大致可以分为两大类,匹配滤波相关器技术和联合变换相关器技术。匹配滤波相关器技术需要制作匹配滤波器,由于制作各种匹配滤波器所需数据的计算量非常大,使得处理精度降低,而且制作滤波器的工艺也不稳定,导致处理信息时的误差,比较不受欢迎;而联合变换相关器不需要制作匹配滤波器,可以实现精确定位,加之使用空间光调制器时可以实现实时自动识别,具有可编程性、灵活性的特点,因而在目标识别领域应用前景较好。近年来,联合变换相关器识别能力的改进主要集中在对图像预处理和对功率谱及相关峰的非线性处理上。
微分算法是一种重要的数字图像处理方法,图像的边缘是图像信息的重要部分,它对图像的识别和理解非常关键,是高级数字图像处理的基础。本文在充分理解微分算法的基础上,首先,将微分算法应用到联合变换相关器中,使得相关输出得到了很好的改善,为联合变换相关器的研究提供了新的方法和途径;其次,将拉普拉斯算法应用到输入图像的功率谱处理,降低了相关峰的零级衍射,获得了较好的相关输出;最后,对相关峰进行了归一化和黑白二值化处理,使得图像相关点的能量相对集中,且亮暗分明,便于识别目标物体。
本文将联合输入图像经过微分处理后输入到空间光调制器,对得到的功率谱和相关峰进行拉普拉斯非线性变换以及归一化和黑白二值化处理,得到了很好的相关输出,增强了相关峰的对比度,使得识别的效率得到提高。
The methods of optical correlation are divided into two categories, matched filter correlator(MFC) and joint transform correlator(JTC). MFC techniques need to make matched filter, The data required for making MFC is very large, This reduces the precessing precision and makes the filtering process unstable,.so the MFC is unpopular; The JTC does not require making a matched filter, and can easily achieve precisely positioning. The use of spatial light modulator(SLM) real-time, automatic programmable and flexibleIn recent years, improvement of identifition ability of JTC is mainly foused on preprocessing of input image,nonlinear processing of power spectrum and correlation peak.
Differential algorithm is an important method of digital image processing. image edge is the important part of image information, and it is critical to image recognition and understanding,It’s also the basis of advanced digital image processing。In this paper, Firstly, the differential algorithm is applied inJTC.The output of JTC was greatly improved and a new path was opend for the development of JTC ; Secondly, the Laplace algorithm was applied to process the power spectrum of input image, The zero-order diffraction of the correlation peak was apparently reduced and a better correlation output had been got; Finally, the correlation peaks were normalized, turned into black and white.and binary, This maked the correlation energy more concentrated and easy to identify the target object.
In this paper, byprocessing the joint input image in spatial light modulator with differential algorithms,processing the power spectrum Laplace algorithm,and processing the correlation output with normalization and binarization method,the contrast of correlation peak was endhanced and the efficencly identification was increased.
微分算法是一种重要的数字图像处理方法,图像的边缘是图像信息的重要部分,它对图像的识别和理解非常关键,是高级数字图像处理的基础。本文在充分理解微分算法的基础上,首先,将微分算法应用到联合变换相关器中,使得相关输出得到了很好的改善,为联合变换相关器的研究提供了新的方法和途径;其次,将拉普拉斯算法应用到输入图像的功率谱处理,降低了相关峰的零级衍射,获得了较好的相关输出;最后,对相关峰进行了归一化和黑白二值化处理,使得图像相关点的能量相对集中,且亮暗分明,便于识别目标物体。
本文将联合输入图像经过微分处理后输入到空间光调制器,对得到的功率谱和相关峰进行拉普拉斯非线性变换以及归一化和黑白二值化处理,得到了很好的相关输出,增强了相关峰的对比度,使得识别的效率得到提高。
The methods of optical correlation are divided into two categories, matched filter correlator(MFC) and joint transform correlator(JTC). MFC techniques need to make matched filter, The data required for making MFC is very large, This reduces the precessing precision and makes the filtering process unstable,.so the MFC is unpopular; The JTC does not require making a matched filter, and can easily achieve precisely positioning. The use of spatial light modulator(SLM) real-time, automatic programmable and flexibleIn recent years, improvement of identifition ability of JTC is mainly foused on preprocessing of input image,nonlinear processing of power spectrum and correlation peak.
Differential algorithm is an important method of digital image processing. image edge is the important part of image information, and it is critical to image recognition and understanding,It’s also the basis of advanced digital image processing。In this paper, Firstly, the differential algorithm is applied inJTC.The output of JTC was greatly improved and a new path was opend for the development of JTC ; Secondly, the Laplace algorithm was applied to process the power spectrum of input image, The zero-order diffraction of the correlation peak was apparently reduced and a better correlation output had been got; Finally, the correlation peaks were normalized, turned into black and white.and binary, This maked the correlation energy more concentrated and easy to identify the target object.
In this paper, byprocessing the joint input image in spatial light modulator with differential algorithms,processing the power spectrum Laplace algorithm,and processing the correlation output with normalization and binarization method,the contrast of correlation peak was endhanced and the efficencly identification was increased.
引文
[1]王仕墦.信息光学理论与应用[M].北京:北京邮电大学出版社,2003:1-282.
[2]宋丰华.现代空间光电信息处理技术及应用[M].北京:国防工业出版社,2004:l-325.
[3]张凤生,王海峰,时洪光.基于CCD的乳腺X线摄影图像预处理方法的研究[J].应用光学,2010,408-411.
[4] WANG Song-wei, LI Yan-jun, Zhang ke.3D Target Recognition Based on Fast Skeleton Extraction[J], ACTA PHOTONICA SINICA,2010,1278-1282.
[5] Vallder Lugst.Singal deteetion by complexs patialfiltering[J].IEEE.Trans.INF.theory.IT.1964,138-144.
[6] C.S.Weaer and J.W.Goodman.A technique for optically convolving two functions[G]. Appl.IPT.1966,6(5):1248-1249.
[7] J.E.Rau.Deteetion of difference in real distributions[J].Opt.1966(56),1248-1249.
[8] U.Efron.ED. Spatial Light Modulator technology[M].Maccel Dekker:New York 1995:45.
[9] C.Warde,And A.D.Fisher.Spatial Light Modulators applications and functional capabilities[J].in Optical Processing,J.L.Honer ED.,Academic Press,1987,478-518.
[10]宋菲君.Spatial Light Modulators[E].北京理工大学博士生演讲稿,2001.
[11]王庆有.CCD应用技术[M].天津大学出版社.2000.
[12]陈前荣,周金鹏,黎全.光学相关识别2f+p系统[J].激光杂志,2003,24(01):32-35.
[13] DannyRoberg,ColinSoutar,B.V.K.vijayaKumar.Optimaltrade-offfilterfortheeorrelation offingerprints[J].opt.Eng.1999,38(1):108-113.
[14] Song hui,wang Wen sheng,et al.research of rotation and scale shift of object in correlation derection [J]. J.changchun institute of optics and fine mechanics, 2003,26(1):39-41
[15]袁祥岩.光学联合变换相关目标识别技术的研究[D].南京:南京理工大学,2003.
[16] Tien-Hsin Chao , HanyingZhou , George Reyes. High-speed optical Object Recognition Processor with massive Holographic Memory[J]. SPIE2002 ,4929:237-243.
[17]陈宇,苗华.梅林变换在光电混合目标探测技术中的应用[J].中国激光,2009,36(2):421-424.
[18]陈方涵.基于小波多尺度积的目标识别[J],光学学报,2009,29(5):1223-1225.
[19]杨坤.Sobel算子在光学相关目标识别中的应用[J],电脑知识与技术,2010,6(15):4245-4246.
[20]郭文强,侯勇严.数字图像处理[M].西安:西安电子科技大学出版社, 2009:1-3.
[21]姚敏.数字图像处理[M].北京:机械工业出版社,2006:2-88.
[22]谢红梅,俞卞章.基于小波变换数据融合的图像边缘检测算法[J].电路与系统学报.2004,9(2):118-121.
[23]朱良等.综合小渡和模糊方法的图像边缘检侧[J].湖南农业大学学报(自然科版).2004,30(2):184-186.
[24]彭玉楼.综合小波和数学形态学的图像边缘检测[J].长沙交通学院学报.2004,20(4):26-29.
[25]DannyRoberg,ColinSoutar,B.v.K.vijayaKumar.Optimaltrade-offfilterfortheeorrelation offingerprints[J].opt.Eng.1999,38(1):108-113.
[26]朱虹.数字图像处理基础[M].北京:科学出版社,2009:6-9.
[27]杨杰.数字图像处理及MATLAB实现[M].北京:电子工业出版社,2010:4-11.
[28]吕乃光.傅里叶光学(第二版)[M].北京:机械工业出版社,2006:319-320.
[29]马北河.实时傅里叶变换在车牌字符识别中的应用[J].广东工业大学学报,2009,26(3):68-71.
[30]吕乃光.傅里叶光学(第二版)[M].北京:机械工业出版社,2006:321-323
[31]龚声蓉.数字图像处理与分析[M].北京:清华大学出版社,2007:51-53.
[32]郭文强,侯勇严.数字图像处理[M].西安:西安电子科技大学出版社,2009:167-168.
[33]刘宏利,邵磊,金飞华等.旋转不变二值化联合变换相关识别方法[J].应用光学,2007,28(4):472-474.
[34]林培秋,王辉.基于液晶空间光调制器的相息图扫描三维成像[J].光电工程,2010,37(3):138-142.
[35]田荣隆,李尚准.光学相关识别系统[J].激光杂志,1990,1(01):52-55.
[36] Vallder Lugt.Singal deteetion by comPlexs Patialfiltering[J].IEEE Trans.Inf.Theory.IT.1964,57:139-145.
[37] Joseph,W.Goodman.傅里叶光学导论[M].北京:电子工业出版社,2006:215-227.
[38]葛宝臻,陈希明,张以漠.基于光电混合联合变换相关的汽车牌照识别方法的实验研究[J],光电子·激光.1999,10(3):248-251.
[2]宋丰华.现代空间光电信息处理技术及应用[M].北京:国防工业出版社,2004:l-325.
[3]张凤生,王海峰,时洪光.基于CCD的乳腺X线摄影图像预处理方法的研究[J].应用光学,2010,408-411.
[4] WANG Song-wei, LI Yan-jun, Zhang ke.3D Target Recognition Based on Fast Skeleton Extraction[J], ACTA PHOTONICA SINICA,2010,1278-1282.
[5] Vallder Lugst.Singal deteetion by complexs patialfiltering[J].IEEE.Trans.INF.theory.IT.1964,138-144.
[6] C.S.Weaer and J.W.Goodman.A technique for optically convolving two functions[G]. Appl.IPT.1966,6(5):1248-1249.
[7] J.E.Rau.Deteetion of difference in real distributions[J].Opt.1966(56),1248-1249.
[8] U.Efron.ED. Spatial Light Modulator technology[M].Maccel Dekker:New York 1995:45.
[9] C.Warde,And A.D.Fisher.Spatial Light Modulators applications and functional capabilities[J].in Optical Processing,J.L.Honer ED.,Academic Press,1987,478-518.
[10]宋菲君.Spatial Light Modulators[E].北京理工大学博士生演讲稿,2001.
[11]王庆有.CCD应用技术[M].天津大学出版社.2000.
[12]陈前荣,周金鹏,黎全.光学相关识别2f+p系统[J].激光杂志,2003,24(01):32-35.
[13] DannyRoberg,ColinSoutar,B.V.K.vijayaKumar.Optimaltrade-offfilterfortheeorrelation offingerprints[J].opt.Eng.1999,38(1):108-113.
[14] Song hui,wang Wen sheng,et al.research of rotation and scale shift of object in correlation derection [J]. J.changchun institute of optics and fine mechanics, 2003,26(1):39-41
[15]袁祥岩.光学联合变换相关目标识别技术的研究[D].南京:南京理工大学,2003.
[16] Tien-Hsin Chao , HanyingZhou , George Reyes. High-speed optical Object Recognition Processor with massive Holographic Memory[J]. SPIE2002 ,4929:237-243.
[17]陈宇,苗华.梅林变换在光电混合目标探测技术中的应用[J].中国激光,2009,36(2):421-424.
[18]陈方涵.基于小波多尺度积的目标识别[J],光学学报,2009,29(5):1223-1225.
[19]杨坤.Sobel算子在光学相关目标识别中的应用[J],电脑知识与技术,2010,6(15):4245-4246.
[20]郭文强,侯勇严.数字图像处理[M].西安:西安电子科技大学出版社, 2009:1-3.
[21]姚敏.数字图像处理[M].北京:机械工业出版社,2006:2-88.
[22]谢红梅,俞卞章.基于小波变换数据融合的图像边缘检测算法[J].电路与系统学报.2004,9(2):118-121.
[23]朱良等.综合小渡和模糊方法的图像边缘检侧[J].湖南农业大学学报(自然科版).2004,30(2):184-186.
[24]彭玉楼.综合小波和数学形态学的图像边缘检测[J].长沙交通学院学报.2004,20(4):26-29.
[25]DannyRoberg,ColinSoutar,B.v.K.vijayaKumar.Optimaltrade-offfilterfortheeorrelation offingerprints[J].opt.Eng.1999,38(1):108-113.
[26]朱虹.数字图像处理基础[M].北京:科学出版社,2009:6-9.
[27]杨杰.数字图像处理及MATLAB实现[M].北京:电子工业出版社,2010:4-11.
[28]吕乃光.傅里叶光学(第二版)[M].北京:机械工业出版社,2006:319-320.
[29]马北河.实时傅里叶变换在车牌字符识别中的应用[J].广东工业大学学报,2009,26(3):68-71.
[30]吕乃光.傅里叶光学(第二版)[M].北京:机械工业出版社,2006:321-323
[31]龚声蓉.数字图像处理与分析[M].北京:清华大学出版社,2007:51-53.
[32]郭文强,侯勇严.数字图像处理[M].西安:西安电子科技大学出版社,2009:167-168.
[33]刘宏利,邵磊,金飞华等.旋转不变二值化联合变换相关识别方法[J].应用光学,2007,28(4):472-474.
[34]林培秋,王辉.基于液晶空间光调制器的相息图扫描三维成像[J].光电工程,2010,37(3):138-142.
[35]田荣隆,李尚准.光学相关识别系统[J].激光杂志,1990,1(01):52-55.
[36] Vallder Lugt.Singal deteetion by comPlexs Patialfiltering[J].IEEE Trans.Inf.Theory.IT.1964,57:139-145.
[37] Joseph,W.Goodman.傅里叶光学导论[M].北京:电子工业出版社,2006:215-227.
[38]葛宝臻,陈希明,张以漠.基于光电混合联合变换相关的汽车牌照识别方法的实验研究[J],光电子·激光.1999,10(3):248-251.