白细胞显微图像识别技术研究
|
摘要
人体白细胞的计数和质量是临床诊断的重要依据。目前国内大多数医院的血液白细胞的检验以人工操作为主，由于受到各种人为因素的影响，使得白细胞的检测质量和效率受到一定的影响。
将计算机图像处理和神经网络理论用于白细胞的检测，可以提高白细胞检测质量和效率。本文研究了用于白细胞识别的图像分析系统的结构和性能，提出了以显微镜、微机和彩色CCD摄像机为主体，应用计算机数字图像等技术实现白细胞分类的实用化系统结构。
本文的主要研究工作是白细胞显微图像分析，综合应用微机控制技术、嵌入式系统技术、数字图像处理技术、小波分析、数学形态学和神经网络理论，围绕着白细胞显微图像分析系统的信息化、自动化程度中的几个关键问题做了以下几方面的研究：1．建立以彩色CCD摄像机成像，计算机进行处理的血细胞自动分类系统模型。2．实现血细胞的定位检出和区域分割。3．对血细胞图像进行数学形态学，彩色光密度和纹理特征的提取。4．运用BP神经网络建立白细胞图像识别分类器，进行网络训练。5．嵌入式Linux显微镜数控平台控制系统设计。本文分别从理论和实际应用的角度对其中的技术难点进行深入分析，在几个方面取得了创新。
采用嵌入式Linux作为开发平台，研究基于Linux的微机控制图像分析系统，利用Linux的开放性和灵活性，精简系统内核，提高了系统的可靠性和稳定性。
基于小波包分解的白细胞图像处理，提出了细胞胞核提取的基本思想，进行了深入的理论研究和实际分析，实现对胞核边缘快速而有效的提取。
应用数学形态学流域分割算法对具有连通或相似灰度的目标图像进行检测与标示，解决了一般数字图像处理技术难以分割具有交叠区域的白细胞图像的问题，在细胞浆的边缘分割中具有创新意义。
分析了图像识别系统的处理流程，设计了基于Linux平台的上下位机两级控制系统，实现了Linux下的接口通信和视频采集，并将系统软件嵌入到硬件中，显著提高控制精度和系统的响应速度。
本课题中的生物显微镜数控平台已经在广州光学仪器厂投入生产。理论分析
广东工业大学工学硕士学位论文
和应用表明，本系统具有实际的应用价值。
The amount and quality of leucocytes is a important gist in the clinical diagnosis. The measure of blood cells in most domestic hospital relies mainly on manual operation at present, It is influenced by various kinds of human factors, so the examination qualities of the leucocytes are partly effected, and the examination efficiency of the leucocytes is effected too.
The application of computer image processing and neural network can greatly improve the analytical efficiency of leucocytes. This paper discussed the structure and performance of automatic classifies system of leucocytes. The system is composed by microscope, microcomputer and chromatic CCD camera, it use the technology of digital image processing and neural network theory to realize automatic classifies of the leucocytes.
The main work of the paper is study on the analysis and the recognition system of leucocytes micrograph. The technology of computer control and embedded, theory of digital image processing, wavelet, mathematics morphology, and neural network pattern-recognition, are applied on the system. Around several key problems, which analytical system of leucocytes micrograph is of information and automatization, have studied on the following aspects: 1.Design automatic classifies system model of blood cell, which is based on imaging of colored CCD camera and computer control. 2.Realize the localization and examined out of the blood cell and the area cutted. 3.Using mathematics morphology on the abstraction of chromatic light density and veins characteristic of the blood cell image. 4.Using BP neural network on setting up recognition classifying device of leucocytes images, and then train it. 5. The design of microscope's NC platform control system Based on the embedded Linux. The paper has analyzed deeply on the diffi
cult point of the theory and practical application, and acquired some creation in the following aspects.
The paper has studied the image analysis system based on the Linux operation
system. Using the open and agility of the Linux we have simplified the system's kernel and improved the system's reliability and stability.
To analyze the leucocytes image by using wavelet, we bring forward the idea of detecting leucocytes' karyon. Through analyzing deeply, we realize the fast-time leucocytes image detection.
In this paper we have used the watershed segmentation arithmetic of mathematics morphologic to analyze and mark the sequential and similar gray target image. As a result, resolved the difficult problem that general technology of digital image can't detect leucocytes' edge of overlapping cells, which has creative meaning in the edge segmentation technology of leucocytes plasm.
The paper discussed the disposed flow of image recognition system, and designed the interface communication of the control system that are master and slave computers based on the Linux platform, then embeds the control programs into the system hardware. All of them can greatly improve the control precision and the response rapidity of system.
Microscope NC platform of this system has been finished. It has putted into production in Guangzhou optical instrument factory. Proved by the analyzing theory and practical application, the system is of applied value.
将计算机图像处理和神经网络理论用于白细胞的检测，可以提高白细胞检测质量和效率。本文研究了用于白细胞识别的图像分析系统的结构和性能，提出了以显微镜、微机和彩色CCD摄像机为主体，应用计算机数字图像等技术实现白细胞分类的实用化系统结构。
本文的主要研究工作是白细胞显微图像分析，综合应用微机控制技术、嵌入式系统技术、数字图像处理技术、小波分析、数学形态学和神经网络理论，围绕着白细胞显微图像分析系统的信息化、自动化程度中的几个关键问题做了以下几方面的研究：1．建立以彩色CCD摄像机成像，计算机进行处理的血细胞自动分类系统模型。2．实现血细胞的定位检出和区域分割。3．对血细胞图像进行数学形态学，彩色光密度和纹理特征的提取。4．运用BP神经网络建立白细胞图像识别分类器，进行网络训练。5．嵌入式Linux显微镜数控平台控制系统设计。本文分别从理论和实际应用的角度对其中的技术难点进行深入分析，在几个方面取得了创新。
采用嵌入式Linux作为开发平台，研究基于Linux的微机控制图像分析系统，利用Linux的开放性和灵活性，精简系统内核，提高了系统的可靠性和稳定性。
基于小波包分解的白细胞图像处理，提出了细胞胞核提取的基本思想，进行了深入的理论研究和实际分析，实现对胞核边缘快速而有效的提取。
应用数学形态学流域分割算法对具有连通或相似灰度的目标图像进行检测与标示，解决了一般数字图像处理技术难以分割具有交叠区域的白细胞图像的问题，在细胞浆的边缘分割中具有创新意义。
分析了图像识别系统的处理流程，设计了基于Linux平台的上下位机两级控制系统，实现了Linux下的接口通信和视频采集，并将系统软件嵌入到硬件中，显著提高控制精度和系统的响应速度。
本课题中的生物显微镜数控平台已经在广州光学仪器厂投入生产。理论分析
广东工业大学工学硕士学位论文
和应用表明，本系统具有实际的应用价值。
The amount and quality of leucocytes is a important gist in the clinical diagnosis. The measure of blood cells in most domestic hospital relies mainly on manual operation at present, It is influenced by various kinds of human factors, so the examination qualities of the leucocytes are partly effected, and the examination efficiency of the leucocytes is effected too.
The application of computer image processing and neural network can greatly improve the analytical efficiency of leucocytes. This paper discussed the structure and performance of automatic classifies system of leucocytes. The system is composed by microscope, microcomputer and chromatic CCD camera, it use the technology of digital image processing and neural network theory to realize automatic classifies of the leucocytes.
The main work of the paper is study on the analysis and the recognition system of leucocytes micrograph. The technology of computer control and embedded, theory of digital image processing, wavelet, mathematics morphology, and neural network pattern-recognition, are applied on the system. Around several key problems, which analytical system of leucocytes micrograph is of information and automatization, have studied on the following aspects: 1.Design automatic classifies system model of blood cell, which is based on imaging of colored CCD camera and computer control. 2.Realize the localization and examined out of the blood cell and the area cutted. 3.Using mathematics morphology on the abstraction of chromatic light density and veins characteristic of the blood cell image. 4.Using BP neural network on setting up recognition classifying device of leucocytes images, and then train it. 5. The design of microscope's NC platform control system Based on the embedded Linux. The paper has analyzed deeply on the diffi
cult point of the theory and practical application, and acquired some creation in the following aspects.
The paper has studied the image analysis system based on the Linux operation
system. Using the open and agility of the Linux we have simplified the system's kernel and improved the system's reliability and stability.
To analyze the leucocytes image by using wavelet, we bring forward the idea of detecting leucocytes' karyon. Through analyzing deeply, we realize the fast-time leucocytes image detection.
In this paper we have used the watershed segmentation arithmetic of mathematics morphologic to analyze and mark the sequential and similar gray target image. As a result, resolved the difficult problem that general technology of digital image can't detect leucocytes' edge of overlapping cells, which has creative meaning in the edge segmentation technology of leucocytes plasm.
The paper discussed the disposed flow of image recognition system, and designed the interface communication of the control system that are master and slave computers based on the Linux platform, then embeds the control programs into the system hardware. All of them can greatly improve the control precision and the response rapidity of system.
Microscope NC platform of this system has been finished. It has putted into production in Guangzhou optical instrument factory. Proved by the analyzing theory and practical application, the system is of applied value.
引文
[1]张之南．血液病诊断及疗效标准[M]．北京：科学出版社，1998．188～178
[2]柯行斌，王汝传，白细胞图像分割的研究与实现，南京邮电学院计算机科学与技术系，南京邮电学院学报，2003．9
[3]张勇，张强，虞烈．真彩色血细胞显微图像自动识别系统研究[J]．西安交通大学学报，1999，33(2)：107～108
[4]Bymc SN, Halliday GM. Dendritic cells. Making progress with tumor regression[J]. Immunol Cell Biol, 2002, 80(6). 520～530
[5]Zhou Y, Bosch ML, Salgaller ML. Current methods for loading denchitic cells with tumor antigen for the induction of antitumor immunity[J]. J Immunothor, 2002, 25(4): 289～303
[6]钟永根，封慰莹，邵剑峰等．恶性血液病患者白细胞数变化与感染关系分析．白血病．淋巴溜，2003，12(3)
[7]张勇，孙岩桦，虞烈．一种有效的白细胞图像彩色空间序贯分割方法．西安交通大学学报，1998，32(8)
[8]陆道墙．白血病治疗学．第1版．北京：科学出版社，1994．84
[9]刘润生，罗建民，费振国．急性白血病与医院感染．中华医院感染学杂志，1994，4(1)
[10]李昕权，范炜，阎安文．高白细胞性白血病患者细胞分离治疗及端粒酶检侧．白血病，2000．9：339
[11]Hiddermann W, Martin W R, 5auerland C M, ei al. Defintion of refaetorineas against conventional chemotherapy in acute myeloidleukemia, a proposal based on the results of retreatment by thioguaint, cytosine arabinoside, and daunorubicin(TAD9) in 150 patients affter standardized fistline therapy[J].l.eukemia. 1990.4:181
[12]吴黎明，周曲，邓耀华等．维氏硬度试验压痕图的小波分析与自动计算硬度．中国机械工程，2004，5(6)：498～500
[13]Castleman K．R10数字图像处理．朱志刚等译．北京：电子工业出版社，1998，30～65
[14]张兆礼．现代图像处理技术及Matlab实现．北京：人民邮电出版社，2001
[15]付忠良．图像阈值选取方法的构造．中国图形图像学报，2000，5(6)
[16]Ming-Hsuan Yang, Tim Persons, Chris Wyatt et al. Survey of Image Segmentation,1998.4.23
[17]张翔，刘媚洁，陈立伟．基于数学形态学的边缘提取方法．电子科技大学学报，2002，31(5)：492～493
[18]Ridella S, Rovetta S. Circular backpropagation networks for classification, IEEE Trans.Neural Networks. 1997,18(1): 84～97
[19]Salah, A.A.; Alpaydin, E.; Akarun, L. A selective attention-based method for visual pattern recognition with application to handwritten digit recognition and face recognition, Pattern Analysis and Machine Intelligence, IEEE Transactions on, Volume:24 Issue:3, March 2002
[20]张广军，李鑫，魏振忠。基于BP网络的结构光三维视觉检测方法研究．仪器仪表学报，2002，23(1)：31～35
[21]刘岚，秦洪．用于神经网络模式识别的一种改进的BP算法．信息技术，2002，(6)：6～7，11
[22]Martin T. Hagan, Howard B.Demuth, Mark H.Beale．神经网络设计．第1版．戴葵等译．北京：机械工业出版社，2002．201～235
[23]Li Zeyu, Tang Shiwei, Wang Hao. Fast recognition of handwritten digits using pairwise coupling support vector machine, Neural Networks, 2002. IJCNN 02. Proceedings of the 2002 International Joint Conference on, 2002, vol.1, Page(s):878～883
[24]高文，计算机视觉——算法与系统原理．北京：清华大学出版社，1999
[25]刘国祥，郭永彩，高潮等．血液细胞免疫特征模式识别技术研究．重庆大学学报，2003，26(1)：109～111，146
[26]陆建峰，杨静宁，唐振民等．重叠细胞图像分离算法的设计[J]．计算机研究与发展，2000，37(2)：228～232
[27]吴黎明，周曲，邓耀华等．基于小波分析的维氏硬度自动测试．理化检验-物
理分册，2004，40(1)：19～22
[28]吴黎明，周曲，邓耀华等．应用数字图像小波多分辨率理论测试维氏硬度．金属热处理，2004，29(6)：55～58
[29]马维祯，殷瑞祥．子波分析与子波变换[M]．广州：华南理工大学出版社，1996．28～31
[30]沈凤麟，陈和晏．生物医学随机信号处理[M]．合肥：中国科技大学出版社，1999．397～399
[31]刘海鹰，黄胜华，彭思龙等．基于小波多尺度分析的图像快速匹配模型[J]．中国图像图形学报，1998，3(11)：907～912
[32]张翔，刘媚洁，陈立伟．基于数学形态学的边缘提取方法．电子科技大学学报，2002，31(5)：492～493
[33]Maragos P.Differential morphology and image processing. IEEE Trans Image Processing, 1996, 5(6): 922～937
[34]崔屹．图像处理与分析-数学形态学方法与应用．北京：科学出版社，2000
[35]吴丹，刘修国，尚建嘎．数学形态学在图像处理与分析中的应用及展望．工程图学学报，2003(2)：121～125
[36]Carnell Hampton, Tim Persons, Chris Wyatt et al. Survey of Image Segmentation, 1998.4.23
[37]汪定慧，程杰，万遂人．基于彩色图像分析的白细胞自动分类算法[J]．东南大学学报(自然科学版)，2000，30(5)：16～20
[38]Paillou P, Gelautz M. Relief reconstruction from SAR stereo pairs-The optimal gradient matching method [J]. IEEE Trans on Geoscience and Remote Sensing, 1999, 37(4): 2099～2107
[39]Soraluze, I.; Rodriguez, C.; Boto, F.; Perez, A. Multidimensional multistage K-NN classifiers for handwritten digit recognition, Frontiers in Handwriting Recognition, 2002. Proceedings. Eighth International Workshop on, 2002, Page(s): 19～23
[40]边肇祺，张学工．模式识别．北京：清华大学出版社，2001．1
[41]吴拥军，吴逸明，屈凌波等．模式识别及人工神经网络技术在肺癌组织分型中的应用研究．计算机与应用化学，2002，19(4)：419～421
[42]吴逸明．职业性肺癌高危人群筛检的聚类与判别分析[J]．系统研究，2002，3
[43]高新波．模糊聚类分析在模式识别中的应用．西安电子科技大学博士论文，1999
[44]李刚，王霄，蔡兰．基于神经网络的控制图模式识别技术研究．制造业自动化，2000，22(5)：31～34
[45]Olivera, L.S.; Benahmed, N.;Sabourin, R.; Bortolozzi, F.; Suen, C.Y. Feature subset selection using genetic algorithms for handwritten digit recognition, Computer Graphics and Image Processing, 2001 Proceedings of XIV Brazilian Symposium on, 2001, Page(s):362～369
[46]邓耀华，吴黎明，唐露新等．实时热处理网络测控系统设计和实现．广东自动化与信息工程，2002，23(1)
[47]Mohammed J Kahir.Redhat Linux 7服务器使用指南[M]．路晓时译．北京：电子工业出版社，2001
[48]钟汉如，王创生．嵌入式Linux的中断处理与实时调度的实现机制．计算机工程，2002．10
[49]Beck M,Bohme H,Dziadzka M,et al.Linux Kernel Internels (Second Edition). Addison-Wesley, 1997
[50]王卫华，陈卫东，顾岳．Linux环境下的实时视频采集．电子技术，2000，(4)：30～32
[2]柯行斌，王汝传，白细胞图像分割的研究与实现，南京邮电学院计算机科学与技术系，南京邮电学院学报，2003．9
[3]张勇，张强，虞烈．真彩色血细胞显微图像自动识别系统研究[J]．西安交通大学学报，1999，33(2)：107～108
[4]Bymc SN, Halliday GM. Dendritic cells. Making progress with tumor regression[J]. Immunol Cell Biol, 2002, 80(6). 520～530
[5]Zhou Y, Bosch ML, Salgaller ML. Current methods for loading denchitic cells with tumor antigen for the induction of antitumor immunity[J]. J Immunothor, 2002, 25(4): 289～303
[6]钟永根，封慰莹，邵剑峰等．恶性血液病患者白细胞数变化与感染关系分析．白血病．淋巴溜，2003，12(3)
[7]张勇，孙岩桦，虞烈．一种有效的白细胞图像彩色空间序贯分割方法．西安交通大学学报，1998，32(8)
[8]陆道墙．白血病治疗学．第1版．北京：科学出版社，1994．84
[9]刘润生，罗建民，费振国．急性白血病与医院感染．中华医院感染学杂志，1994，4(1)
[10]李昕权，范炜，阎安文．高白细胞性白血病患者细胞分离治疗及端粒酶检侧．白血病，2000．9：339
[11]Hiddermann W, Martin W R, 5auerland C M, ei al. Defintion of refaetorineas against conventional chemotherapy in acute myeloidleukemia, a proposal based on the results of retreatment by thioguaint, cytosine arabinoside, and daunorubicin(TAD9) in 150 patients affter standardized fistline therapy[J].l.eukemia. 1990.4:181
[12]吴黎明，周曲，邓耀华等．维氏硬度试验压痕图的小波分析与自动计算硬度．中国机械工程，2004，5(6)：498～500
[13]Castleman K．R10数字图像处理．朱志刚等译．北京：电子工业出版社，1998，30～65
[14]张兆礼．现代图像处理技术及Matlab实现．北京：人民邮电出版社，2001
[15]付忠良．图像阈值选取方法的构造．中国图形图像学报，2000，5(6)
[16]Ming-Hsuan Yang, Tim Persons, Chris Wyatt et al. Survey of Image Segmentation,1998.4.23
[17]张翔，刘媚洁，陈立伟．基于数学形态学的边缘提取方法．电子科技大学学报，2002，31(5)：492～493
[18]Ridella S, Rovetta S. Circular backpropagation networks for classification, IEEE Trans.Neural Networks. 1997,18(1): 84～97
[19]Salah, A.A.; Alpaydin, E.; Akarun, L. A selective attention-based method for visual pattern recognition with application to handwritten digit recognition and face recognition, Pattern Analysis and Machine Intelligence, IEEE Transactions on, Volume:24 Issue:3, March 2002
[20]张广军，李鑫，魏振忠。基于BP网络的结构光三维视觉检测方法研究．仪器仪表学报，2002，23(1)：31～35
[21]刘岚，秦洪．用于神经网络模式识别的一种改进的BP算法．信息技术，2002，(6)：6～7，11
[22]Martin T. Hagan, Howard B.Demuth, Mark H.Beale．神经网络设计．第1版．戴葵等译．北京：机械工业出版社，2002．201～235
[23]Li Zeyu, Tang Shiwei, Wang Hao. Fast recognition of handwritten digits using pairwise coupling support vector machine, Neural Networks, 2002. IJCNN 02. Proceedings of the 2002 International Joint Conference on, 2002, vol.1, Page(s):878～883
[24]高文，计算机视觉——算法与系统原理．北京：清华大学出版社，1999
[25]刘国祥，郭永彩，高潮等．血液细胞免疫特征模式识别技术研究．重庆大学学报，2003，26(1)：109～111，146
[26]陆建峰，杨静宁，唐振民等．重叠细胞图像分离算法的设计[J]．计算机研究与发展，2000，37(2)：228～232
[27]吴黎明，周曲，邓耀华等．基于小波分析的维氏硬度自动测试．理化检验-物
理分册，2004，40(1)：19～22
[28]吴黎明，周曲，邓耀华等．应用数字图像小波多分辨率理论测试维氏硬度．金属热处理，2004，29(6)：55～58
[29]马维祯，殷瑞祥．子波分析与子波变换[M]．广州：华南理工大学出版社，1996．28～31
[30]沈凤麟，陈和晏．生物医学随机信号处理[M]．合肥：中国科技大学出版社，1999．397～399
[31]刘海鹰，黄胜华，彭思龙等．基于小波多尺度分析的图像快速匹配模型[J]．中国图像图形学报，1998，3(11)：907～912
[32]张翔，刘媚洁，陈立伟．基于数学形态学的边缘提取方法．电子科技大学学报，2002，31(5)：492～493
[33]Maragos P.Differential morphology and image processing. IEEE Trans Image Processing, 1996, 5(6): 922～937
[34]崔屹．图像处理与分析-数学形态学方法与应用．北京：科学出版社，2000
[35]吴丹，刘修国，尚建嘎．数学形态学在图像处理与分析中的应用及展望．工程图学学报，2003(2)：121～125
[36]Carnell Hampton, Tim Persons, Chris Wyatt et al. Survey of Image Segmentation, 1998.4.23
[37]汪定慧，程杰，万遂人．基于彩色图像分析的白细胞自动分类算法[J]．东南大学学报(自然科学版)，2000，30(5)：16～20
[38]Paillou P, Gelautz M. Relief reconstruction from SAR stereo pairs-The optimal gradient matching method [J]. IEEE Trans on Geoscience and Remote Sensing, 1999, 37(4): 2099～2107
[39]Soraluze, I.; Rodriguez, C.; Boto, F.; Perez, A. Multidimensional multistage K-NN classifiers for handwritten digit recognition, Frontiers in Handwriting Recognition, 2002. Proceedings. Eighth International Workshop on, 2002, Page(s): 19～23
[40]边肇祺，张学工．模式识别．北京：清华大学出版社，2001．1
[41]吴拥军，吴逸明，屈凌波等．模式识别及人工神经网络技术在肺癌组织分型中的应用研究．计算机与应用化学，2002，19(4)：419～421
[42]吴逸明．职业性肺癌高危人群筛检的聚类与判别分析[J]．系统研究，2002，3
[43]高新波．模糊聚类分析在模式识别中的应用．西安电子科技大学博士论文，1999
[44]李刚，王霄，蔡兰．基于神经网络的控制图模式识别技术研究．制造业自动化，2000，22(5)：31～34
[45]Olivera, L.S.; Benahmed, N.;Sabourin, R.; Bortolozzi, F.; Suen, C.Y. Feature subset selection using genetic algorithms for handwritten digit recognition, Computer Graphics and Image Processing, 2001 Proceedings of XIV Brazilian Symposium on, 2001, Page(s):362～369
[46]邓耀华，吴黎明，唐露新等．实时热处理网络测控系统设计和实现．广东自动化与信息工程，2002，23(1)
[47]Mohammed J Kahir.Redhat Linux 7服务器使用指南[M]．路晓时译．北京：电子工业出版社，2001
[48]钟汉如，王创生．嵌入式Linux的中断处理与实时调度的实现机制．计算机工程，2002．10
[49]Beck M,Bohme H,Dziadzka M,et al.Linux Kernel Internels (Second Edition). Addison-Wesley, 1997
[50]王卫华，陈卫东，顾岳．Linux环境下的实时视频采集．电子技术，2000，(4)：30～32
© 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 |