嵌入式液体杂质检测系统研究
|
摘要
医用液体在生产过程中混入的杂质严重危害了使用者的生命健康。目前采用的人工检测方法受灯检工主观性影响而导致检测效果不够理想,研究医用液体杂质自动检测系统具有重要的理论意义和应用价值。
本文首先介绍了液体杂质检测技术的研究背景和研究现状,并对涉及的理论与技术进行了探讨。在分析数字图像处理技术常用算法的基础上,针对杂质图像的特征,着重介绍了与液体杂质检测系统相关的图像增强、图像分割和二值数学形态学等图像处理技术。
接着,介绍了脉冲耦合神经网络,分析了其基本模型和简化模型。针对模型中参数需要重新设定的缺陷,研究了自适应的脉冲耦合神经网络,并将最小二乘法准则和梯度下降法融合其中,解决了脉冲耦合神经网络点火时间序列对光照敏感度的需求。在此基础上,结合数字图像处理的相关理论,研究了基于脉冲耦合神经网络的图像增强、图像分割、区域标记和面积计算方法。
最后,设计了以DM642为核心处理器的嵌入式杂质检测系统,并对系统中的硬件和软件分别进行了详细的分析和设计。系统硬件主要包括视频采集、图像处理和控制单元三个部分。软件实现了视频采集、图像增强、杂质检测、杂质识别和不合格品剔除等功能,并进行了实验测试,实验结果证明了所提算法和设计的系统满足在线检测要求。
The impurities which mixed in the medical liquid in the production process has seriously harm to the health of patients. As the result of the manual detection methods which affected by the subjectivity is not satisfactory, therefore the research of automatic detection system for medical liquid impurity has great theoretical significance and application value.
Firstly, the related theory and technology of the detection of liquid impurities are introduced. For the characteristic of impurities image, the digital image processing technology and commonly used algorithms which related to the detection system of liquid impurities were analyzed, such as image enhancement, image segmentation and binary image with mathematical morphology.
Secondly, the basic model and the simplified model of the pulse coupled neural network were analyzed. As the shortcomings of parameters need to be re-set in the model, the least-squares approach and gradient descent algorithm have been integrated in the adaptive pulse coupled neural network, which solved the demand of ignition time series to the light sensitivity. On this basis, the image enhancement, image segmentation, region labeling algorithm and area calculation method based on pulse coupled neural network which combined with the relevant theory of digital image processing have been researched.
Finally, the impurity detection embedded system has been designed based on the processor DM642, and the hardware and software analysis and design of the system have been given separately in detail. The system hardware includes three parts:video capture, image processing and control unit, and the following functions which include system video capture, image enhancement, impurity detection, impurity identification and substandard products removal have been implemented in software. The experimental results show that the proposed algorithm and designed system meets the online detection requirements.
本文首先介绍了液体杂质检测技术的研究背景和研究现状,并对涉及的理论与技术进行了探讨。在分析数字图像处理技术常用算法的基础上,针对杂质图像的特征,着重介绍了与液体杂质检测系统相关的图像增强、图像分割和二值数学形态学等图像处理技术。
接着,介绍了脉冲耦合神经网络,分析了其基本模型和简化模型。针对模型中参数需要重新设定的缺陷,研究了自适应的脉冲耦合神经网络,并将最小二乘法准则和梯度下降法融合其中,解决了脉冲耦合神经网络点火时间序列对光照敏感度的需求。在此基础上,结合数字图像处理的相关理论,研究了基于脉冲耦合神经网络的图像增强、图像分割、区域标记和面积计算方法。
最后,设计了以DM642为核心处理器的嵌入式杂质检测系统,并对系统中的硬件和软件分别进行了详细的分析和设计。系统硬件主要包括视频采集、图像处理和控制单元三个部分。软件实现了视频采集、图像增强、杂质检测、杂质识别和不合格品剔除等功能,并进行了实验测试,实验结果证明了所提算法和设计的系统满足在线检测要求。
The impurities which mixed in the medical liquid in the production process has seriously harm to the health of patients. As the result of the manual detection methods which affected by the subjectivity is not satisfactory, therefore the research of automatic detection system for medical liquid impurity has great theoretical significance and application value.
Firstly, the related theory and technology of the detection of liquid impurities are introduced. For the characteristic of impurities image, the digital image processing technology and commonly used algorithms which related to the detection system of liquid impurities were analyzed, such as image enhancement, image segmentation and binary image with mathematical morphology.
Secondly, the basic model and the simplified model of the pulse coupled neural network were analyzed. As the shortcomings of parameters need to be re-set in the model, the least-squares approach and gradient descent algorithm have been integrated in the adaptive pulse coupled neural network, which solved the demand of ignition time series to the light sensitivity. On this basis, the image enhancement, image segmentation, region labeling algorithm and area calculation method based on pulse coupled neural network which combined with the relevant theory of digital image processing have been researched.
Finally, the impurity detection embedded system has been designed based on the processor DM642, and the hardware and software analysis and design of the system have been given separately in detail. The system hardware includes three parts:video capture, image processing and control unit, and the following functions which include system video capture, image enhancement, impurity detection, impurity identification and substandard products removal have been implemented in software. The experimental results show that the proposed algorithm and designed system meets the online detection requirements.
引文
[1]刘俊敏,医学图像处理技术的现状和发展方向[J],医疗卫生设备,Vol.26,No.12,2005.
[2]国家药典委员会,中华人民共和国[M].2版.北京：化学工业出版社,2005.
[3]E.N.Malamas,E-GPetrakis,M.Zervakis,etal.A survey on industrial vision system s[J],applications and tools,2003,21(2):171-188.
[4]K.Miehael,Cheezum,F.EWilliam,etal.Quantitative ComParison of Algorithms fo r Tracking Single Fluoreseent Particles[J].BioPhysical Journal,2001,81(10):237 8-2388.
[5]杨福刚,孙同景,宋松林.基于人工免疫算法的弱小目标检测方法[J].电子测量与仪器学报,2008,22(1)：20-24.
[6]杨福刚,孙同景,宋松林.基于机器视觉的灯检机关键技术研究[J].仪器仪表学报,2008,29(3)：562-566.
[7]章毓晋.图像工程(上册)[M].北京：清华大学出版社,1999.
[8]李弼程等.智能图像处理技术[M].北京：电子工业出版社,2004.07.
[9]Anil K Jain.数字图像处理技术[M].韩博,徐枫译.北京：清华大学出版社,2006.
[10]薛景浩,章毓晋,林行刚.基于最大类间后验交叉熵的阈值分割[J].中国图像图形学报,1999,4(2)：110-114.
[11]Li C H,Lee C K.Minimum cross-entropy thresholding[J].Pattern Recognition, 1993,26(14):617-625.
[12]Pal N R.On minimum cross-entropy thresholding[J].Pattern Recognition,1996, 29(4):575-580.
[13]Brink A D,Pendock N E.minimum cross-entropy threshold selection[J].Pattern Recognition,1996,29(1):179-188.
[14]Kenneth R Castleman.Digital Image Processing Printice-Hall,Inc[M].北京：清华大学出版社,1988.
[15]Kenneth R Castleman[美].数字图像处理[M].朱志刚等译.北京：电子工业出版社,1998.
[16]Eckhorn R,Reitboeck H J,Arndt M,Dicke P W.Feature linking via stimulus-evoked oscillations:Expermental results from cat visual cortex and functionl implications form a network model[J].Neural Networks,IJCNN,1989,6(1):723-730.
[17]Eckhorn R,Reitboeck H J,Arndt M,Dicke P W.Feature linking via synchoroni zation among distributed assemblies:Simulation results from cat cortex[J].Neu ral Computation,1990,(2):293-307.
[18]马义德,齐春亮.基于遗传算法的脉冲耦合神经网络自动系统的研究[J].系统仿真学报,2006,18(03)：722-724.
[19]马义德,绽琨,齐春亮.自适应脉冲耦合神经网络在图像处理中的应用[J].系统仿真学报,2008,20(11)：2897-2900,2930.
[20]Randy Paul Broussard.Physiologically based vision modeling applications and gradient decent based parameter adaptation of puke coupled neural network s[J].Air Farce Institute of Technology,AFIT/DSG/ENG/96-3M,1997:69-80.
[21]Zhan Kun,ZhaoBin Wang,Ma Yide.The relationship between human visual characterstics and PCNN for image processing[c].IEEE Transactions on Image Processing,2008.
[22]Ma yide,Qi Chunliang.Region labeling method based on double PCNN and morphology[C].International Symposium on Communications and Information Technologies,2007:526-530.
[23]马义德,李廉,戴若兰等.一种基于脉冲耦合神经网络的植物胚性细胞图像分割[J].科学通报,2001,46(21)：1781-1786.
[24]马义德,李廉,戴若兰等.一种基于脉冲耦合神经网络和图像熵的自动图像分割方法[J].通信学报,2002,23(1)：46-51.
[25]岑曙炜.图像骨架和终极腐蚀的若干性质[J].中国图像图形学报,2001,14(1)：99-103.
[26]TMS320DM642 Video/Imaging Fixed-Point Digital Signal Processor Data Sh eet[Z].TI.Inc.,2007
[27]TMS320DM642 Hardware Designers Resource Guide[Z].Tl.Inc.,2005.
[28]盛翠霞,张涛,纪晶,姚清华等.高分辨率CCD芯片FTF4O52M的驱动系统设计[J].光学精密工程,2007,15(4)：564-569.
[29]刘光林,杨世洪,吴钦章等.高分辨率全帧CCD相机电路系统的设计[J].中国科学院研究生学报,2007,24(3)：32-324.
[30]廖红华.基于高分辨率面阵CCD的图像采集系统研究[D].硕士学位论文,重庆：重庆大学,2006.
[31]江思敏,刘畅.TMs32oc6000DSP应用开发教程[M].北京：机械工业出版社,2005.
[32]DSP/BIOS Driver Developer's Guide[Z].Tl.Inc.,2005.
[33]The TMS320DM642 Video Port Mini-Driver[Z].Tl.Inc.,2005.
[2]国家药典委员会,中华人民共和国[M].2版.北京：化学工业出版社,2005.
[3]E.N.Malamas,E-GPetrakis,M.Zervakis,etal.A survey on industrial vision system s[J],applications and tools,2003,21(2):171-188.
[4]K.Miehael,Cheezum,F.EWilliam,etal.Quantitative ComParison of Algorithms fo r Tracking Single Fluoreseent Particles[J].BioPhysical Journal,2001,81(10):237 8-2388.
[5]杨福刚,孙同景,宋松林.基于人工免疫算法的弱小目标检测方法[J].电子测量与仪器学报,2008,22(1)：20-24.
[6]杨福刚,孙同景,宋松林.基于机器视觉的灯检机关键技术研究[J].仪器仪表学报,2008,29(3)：562-566.
[7]章毓晋.图像工程(上册)[M].北京：清华大学出版社,1999.
[8]李弼程等.智能图像处理技术[M].北京：电子工业出版社,2004.07.
[9]Anil K Jain.数字图像处理技术[M].韩博,徐枫译.北京：清华大学出版社,2006.
[10]薛景浩,章毓晋,林行刚.基于最大类间后验交叉熵的阈值分割[J].中国图像图形学报,1999,4(2)：110-114.
[11]Li C H,Lee C K.Minimum cross-entropy thresholding[J].Pattern Recognition, 1993,26(14):617-625.
[12]Pal N R.On minimum cross-entropy thresholding[J].Pattern Recognition,1996, 29(4):575-580.
[13]Brink A D,Pendock N E.minimum cross-entropy threshold selection[J].Pattern Recognition,1996,29(1):179-188.
[14]Kenneth R Castleman.Digital Image Processing Printice-Hall,Inc[M].北京：清华大学出版社,1988.
[15]Kenneth R Castleman[美].数字图像处理[M].朱志刚等译.北京：电子工业出版社,1998.
[16]Eckhorn R,Reitboeck H J,Arndt M,Dicke P W.Feature linking via stimulus-evoked oscillations:Expermental results from cat visual cortex and functionl implications form a network model[J].Neural Networks,IJCNN,1989,6(1):723-730.
[17]Eckhorn R,Reitboeck H J,Arndt M,Dicke P W.Feature linking via synchoroni zation among distributed assemblies:Simulation results from cat cortex[J].Neu ral Computation,1990,(2):293-307.
[18]马义德,齐春亮.基于遗传算法的脉冲耦合神经网络自动系统的研究[J].系统仿真学报,2006,18(03)：722-724.
[19]马义德,绽琨,齐春亮.自适应脉冲耦合神经网络在图像处理中的应用[J].系统仿真学报,2008,20(11)：2897-2900,2930.
[20]Randy Paul Broussard.Physiologically based vision modeling applications and gradient decent based parameter adaptation of puke coupled neural network s[J].Air Farce Institute of Technology,AFIT/DSG/ENG/96-3M,1997:69-80.
[21]Zhan Kun,ZhaoBin Wang,Ma Yide.The relationship between human visual characterstics and PCNN for image processing[c].IEEE Transactions on Image Processing,2008.
[22]Ma yide,Qi Chunliang.Region labeling method based on double PCNN and morphology[C].International Symposium on Communications and Information Technologies,2007:526-530.
[23]马义德,李廉,戴若兰等.一种基于脉冲耦合神经网络的植物胚性细胞图像分割[J].科学通报,2001,46(21)：1781-1786.
[24]马义德,李廉,戴若兰等.一种基于脉冲耦合神经网络和图像熵的自动图像分割方法[J].通信学报,2002,23(1)：46-51.
[25]岑曙炜.图像骨架和终极腐蚀的若干性质[J].中国图像图形学报,2001,14(1)：99-103.
[26]TMS320DM642 Video/Imaging Fixed-Point Digital Signal Processor Data Sh eet[Z].TI.Inc.,2007
[27]TMS320DM642 Hardware Designers Resource Guide[Z].Tl.Inc.,2005.
[28]盛翠霞,张涛,纪晶,姚清华等.高分辨率CCD芯片FTF4O52M的驱动系统设计[J].光学精密工程,2007,15(4)：564-569.
[29]刘光林,杨世洪,吴钦章等.高分辨率全帧CCD相机电路系统的设计[J].中国科学院研究生学报,2007,24(3)：32-324.
[30]廖红华.基于高分辨率面阵CCD的图像采集系统研究[D].硕士学位论文,重庆：重庆大学,2006.
[31]江思敏,刘畅.TMs32oc6000DSP应用开发教程[M].北京：机械工业出版社,2005.
[32]DSP/BIOS Driver Developer's Guide[Z].Tl.Inc.,2005.
[33]The TMS320DM642 Video Port Mini-Driver[Z].Tl.Inc.,2005.