基于视频的动物运动跟踪分析系统及应用研究
|
摘要
行走是一种常见的生物体的整体运动。运动分析是对生物体行走方式的一种检查方法,旨在通过运动学手段,揭示步态异常的环节和影响因素,从而指导康复评定和治疗,也有助于临床诊断和机制研究等。
本论文以国家自然科学基金为背景,针对动物实验研制一套基于视频的低成本运动跟踪分析系统,包括硬件平台和软件设计两大部分,其中硬件平台由高速摄像机、千兆网采集卡和PC机组成;系统软件采用MATLAB编程,完成对视频中实验动物特定关节标识的自动跟踪及运动学参数计算等功能。
本文首先介绍了当前国内外运动跟踪系统的分类,比较应用于运动跟踪的不同机动目标模型、预测算法和目标识别算法等,分析各自的优缺点,然后介绍运动跟踪系统的硬件组成和三维标识的制作过程,详细叙述了跟踪算法的实现。根据大鼠后肢运动强弱机动混合的特点,分别采用了多项式拟合和基于“当前”机动目标模型的改进卡尔曼滤波对运动目标进行位置预测,然后采用数据融合的方法得到最终的预测结果;根据标识大小、标识与背景对比强烈程度的不同,分别选取了峰值跟踪、基于灰度的模板匹配和基于颜色的模板匹配等目标识别方法,最后用误匹配检测判断所识别目标是否正确。此外,本文还对运动学参数和肌电信号参数进行同步后用于评价康复效果,并对脊髓损伤大鼠分别在有、无跑台训练的条件下对康复效果进行运动分析,比较其运动参数,结果表明跑台训练方法对大鼠行走能力的康复有促进作用,该方法可用于脊髓损伤大鼠后肢运动能力的恢复。
Walking is a common movement of animals. Motion analysis is to check the gait pattern, where the kinematic analysis is used to reveal the abnormal gait and the influential factors. Therefore, motion analysis can be used for the evaluation of rehabilitation therapy, the clinical diagnosis and the mechanism research, etc.
This thesis is in part supported by National Natural Science Foundation Project.A video-based motor tracking and analysis system for animal experiments is presented, which includes hardware system and software design.The hardware system consists of a high-speed video camera, a gigabit Ethernet data acquisition card and a personal computer; In the software programming,MATLAB provides the interface to achieve the automatic tracking of specific markers in the video and compute the kinematic parameters.
Firstly, this thesis introduces the classification of motion tracking systems and describes the advantages and disadvantages of some maneuvering target models, motion prediction and target recognition algorithms. Then the hardware configuration of the motion analysis system and the fabrication of three-dimension markers are introduced. After that, this thesis describes the motion tracking algorithms in detail. Considering the coexistance of high and weak maneuverabilityof the limb motion of adult rats, the polynomial fitting and improved Kalman filtering algorithms based on the current model are separately used to predict the preliminary positions of moving targets. Then the resulting predictions are obtained by the data fusion on the preliminary positions. For the markers with different size and different contrast to the background, hot-spot matching, intensity-based and color-based algorithms are used to find the positions of tracked markers; The mismatch detection is applied to enable a motion tracking with reliability. Besides, this thesis describes a method for synchronizing the kinematic data and the electromyography signal. At the end of this thesis, the motion analysis is performed on the normal and incomplete spinal cord injury rats with and without the treadmilltraining . The results show that the treadmill training can improve the weight bearing of the hindlimb of rats with incomplete spinal cord injury.
本论文以国家自然科学基金为背景,针对动物实验研制一套基于视频的低成本运动跟踪分析系统,包括硬件平台和软件设计两大部分,其中硬件平台由高速摄像机、千兆网采集卡和PC机组成;系统软件采用MATLAB编程,完成对视频中实验动物特定关节标识的自动跟踪及运动学参数计算等功能。
本文首先介绍了当前国内外运动跟踪系统的分类,比较应用于运动跟踪的不同机动目标模型、预测算法和目标识别算法等,分析各自的优缺点,然后介绍运动跟踪系统的硬件组成和三维标识的制作过程,详细叙述了跟踪算法的实现。根据大鼠后肢运动强弱机动混合的特点,分别采用了多项式拟合和基于“当前”机动目标模型的改进卡尔曼滤波对运动目标进行位置预测,然后采用数据融合的方法得到最终的预测结果;根据标识大小、标识与背景对比强烈程度的不同,分别选取了峰值跟踪、基于灰度的模板匹配和基于颜色的模板匹配等目标识别方法,最后用误匹配检测判断所识别目标是否正确。此外,本文还对运动学参数和肌电信号参数进行同步后用于评价康复效果,并对脊髓损伤大鼠分别在有、无跑台训练的条件下对康复效果进行运动分析,比较其运动参数,结果表明跑台训练方法对大鼠行走能力的康复有促进作用,该方法可用于脊髓损伤大鼠后肢运动能力的恢复。
Walking is a common movement of animals. Motion analysis is to check the gait pattern, where the kinematic analysis is used to reveal the abnormal gait and the influential factors. Therefore, motion analysis can be used for the evaluation of rehabilitation therapy, the clinical diagnosis and the mechanism research, etc.
This thesis is in part supported by National Natural Science Foundation Project.A video-based motor tracking and analysis system for animal experiments is presented, which includes hardware system and software design.The hardware system consists of a high-speed video camera, a gigabit Ethernet data acquisition card and a personal computer; In the software programming,MATLAB provides the interface to achieve the automatic tracking of specific markers in the video and compute the kinematic parameters.
Firstly, this thesis introduces the classification of motion tracking systems and describes the advantages and disadvantages of some maneuvering target models, motion prediction and target recognition algorithms. Then the hardware configuration of the motion analysis system and the fabrication of three-dimension markers are introduced. After that, this thesis describes the motion tracking algorithms in detail. Considering the coexistance of high and weak maneuverabilityof the limb motion of adult rats, the polynomial fitting and improved Kalman filtering algorithms based on the current model are separately used to predict the preliminary positions of moving targets. Then the resulting predictions are obtained by the data fusion on the preliminary positions. For the markers with different size and different contrast to the background, hot-spot matching, intensity-based and color-based algorithms are used to find the positions of tracked markers; The mismatch detection is applied to enable a motion tracking with reliability. Besides, this thesis describes a method for synchronizing the kinematic data and the electromyography signal. At the end of this thesis, the motion analysis is performed on the normal and incomplete spinal cord injury rats with and without the treadmilltraining . The results show that the treadmill training can improve the weight bearing of the hindlimb of rats with incomplete spinal cord injury.
引文
[1] Chad V. Anderson,Andrew J. Fuglevand.Probability-Based Prediction of Activity in Multiple Arm Muscles:Implications for Functional Electrical Stimulation. Journal of Neurophysiology, 2008, 100(1): 482~494.
[2] Chao Y.S. Justification of travail goniometric for measurement of joint rotation. J Biomech, 1980, 13(12): 989~1006.
[3] Clarke RL,Smith RF and Justesen DR.An infrared device for detecting locomotor activity. Behav Res Methods Instrum Comput, 1985, 17(5): 519~525.
[4] Martin PH and Unwin DM. A microwave Doppler radar activity monitor. Behav Res Methods Instrum, 1980, 12(5):517~520.
[5] Giansanti D, Maccioni G, Macellari V. The development and test of a device for the reconstruction of 3-D position and orientation by means of a kinematic sensor assembly with rate gyroscopes and accelerometers. IEEE Trans Biomed Eng, 2005, 52(3): 1271~1277.
[6] Georgopoulos AP, Schwartz AB, Kettner RE. Neuronal population coding of movement direction. Science, 1986, 233(4771): 1416~1419.
[7] M. Furniss. Motion capture. MIT Communication Forum, 1999.
[8] C. Grow, I. Gordon, R.D. Stuart, A. Adalja. Motion Capture as a Means for Data Acquisition .ARA Focus Group Spring, 1998.
[9] Macmillan DL. A physiological analysis of walking in the American lobster (Homarus americanus), Philos Trans R Soc Lond B Biol Sci , 1975, 270(901): 1~59.
[10] Uter TG. A real-time video system for tracking one- dimensional movements of two objects. IEEE Trans Biomed Eng, 1977, 24(1): 8~75.
[11] Pascual J.Figueroa , Neucimar J. Leite, and Ricardo M.L.Barros. A ?exible software for tracking of markers used in human motion analysis. Computer Methods and Programs in Biomedicine, 2003, 72(2): 155~165.
[12]程建,王东泉,施鹏飞等.基于图象技术的三维步态运动测量.上海交通大学学报, 1998, 32(4): 59~64.
[13]胡琴,王文中,夏时洪等.基于多摄像机的人体步态跟踪方法.计算机工程, 2008, 34(22): 220~222.
[14]王人成,黄昌华,王季军等.基于普通摄像机的人体运动信息检测系统.生物医学工程学杂志, 1999, 16 (4) : 448~452.
[15]王人成,董华,黄昌华等.低成本人体步态在线检测系统.清华大学学报, 2002, 42(2): 165~167.
[16]蔡付文,王人成,李广庆等.低成本人体步态分析系统的研制.康复医学工程, 2008, 23(1): 49~53.
[17] Haritaoglu I,Harwood D, Davis L Y. Real-time surveillance of people and their activities. IEEE Trans Pattern Analysis and Machine Intelligence, 2000, 22(8): 809~830 .
[18]李建华,孙怡,马小妹等.基于距离图像的人体下肢关节点运动跟踪.全国信号处理学术年会论文集, 2001: 333~336.
[19]冯莲,邹北骥,刘相滨等.无标志的人体步行腿部骨架检测与跟踪.计算机工程与科学, 2007, 29(1): 62~65.
[20]潘泉,梁彦,杨峰等.现代目标跟踪与信息融合.北京:国防工业出版社, 2009. 32~39
[21]彭亮.机动目标跟踪算法的研究:[硕士学位论文].华中科技大学图书馆, 2007.
[22]周宏仁,敬忠良,王培德.机动目标跟踪.北京:国防工业出版社, 1991. 10~142
[23] Moose R L. An adaptive state estimation solution to the maneuvering target problem. IEEE Trans. Autom. Control, 1975, AC-20(6): 359~362.
[24] Morgan D R. A target trajectory noise model for Kalman trackers. IEEE Transaction on Aerospace and Electronic Systems, 1976, 12(3): 405~408.
[25]杨耿,和卫星.运动目标图像识别与跟踪系统的研究.计算机测量与控制, 2005, 13(3): 267-269.
[26] Greg Welch, Gary Bishop. An Introduction to the Kalman Filter. SIGGRAPH, USA , 2001, Course 8:20~24.
[27]李尊民.电视图像自动跟踪的基本原理.北京:国防工业出版社, 1998. 41~74
[28]张庆生,邓兵,汪遵懋.低空高射武器观瞄系统的图像处理技术.光学技术,2001,27(5):262~265.
[29]刘宝生,闫莉萍,周东华.几种种经典相似性度量的比较研究.计算机应用研究, 2006, 11: 1~3.
[30]恽晓平.康复疗法评定学.北京:华夏出版社,2005.263~287
[31] Karim Fouad, Gerlinde .S. Metz, Doron Merkler etal. Treadmill training in incomplete spinal cord injured rats. Behavioural Brain Research,2000, 115(1):107~113.
[32] K.G. Pearson, H. Acharya, K. Fouad. A new electrode configuration for recordingelectromyographic activity in behaving mice. Journal of Neuroscience Methods, 2005, 148(1):36~42.
[33] A.J. Spink, R.A.J. Tegelenbosch, M.O.S. Buma,et al.The EthoVision video tracking system—A tool for behavioral phenotyping of transgenic mice.Physiology & Behavior,2001, 73(5):731~744.
[34] Chad V. Anderson, Andrew J. Fuglevand. Probability-Based Prediction of Activity in Multiple Arm Muscles:Implications for Functional Electrical Stimulation. Journal of Neurophysiol, 2008, 100(1): 482~494.
[35]贺兴华,周媛媛,王继阳等. MATLAB 7.X图像处理.北京:人民邮电出版社, 2006. 2~9
[36]杨青智,刘明治.机动目标跟踪技术的研究:[硕士学位论文].华中科技大学图书馆,2008.
[37]方青,梅晓春,张育平.用于机动目标跟踪的Kalman滤波器的设计.雷达科学与技术, 2006, 2(1): 50-55.
[38] Blair W D.Rice T R, Mcdole B S. Least-squares approach to asynchronous data fusion. SPIE Acquisition, Tracking and Pointing VI, USA, 1992, 1697(6):130~139.
[39]章毓晋.图像工程(上册).北京:清华大学出版社, 2001, 82~90.
[40]飞思科技产品研发中心. MATLAB 6.5辅助图像处理.北京:电子工业出版社, 2003. 59~69.
[39] Ian D. Peikona, Nathan A. Fitzsimmonsb, Mikhail A. Lebedev etal. Three-dimensional, automated,real-time video system for tracking limb motion in brain-machine interface studies. Journal of Neuroscience Methods. 2009, 180(2): 224~233.
[42] Gerald L. Scheirman and Phillip J. Cheetham. Motion measurement using the Peak Performance Technologies system. SPIE Image-Based Motion Measurement, 1990, 1356(11): 63~70.
[43] Yury P. Gerasimenko, Ronaldo M. Ichiyama,Igor A etal. Lavrov,Epidural Spinal Cord Stimulation Plus Quipazine Administration Enable Stepping in Complete Spinal Adult Rats. J Neurophysiol, 2007, 98(5): 2525~2536.
[2] Chao Y.S. Justification of travail goniometric for measurement of joint rotation. J Biomech, 1980, 13(12): 989~1006.
[3] Clarke RL,Smith RF and Justesen DR.An infrared device for detecting locomotor activity. Behav Res Methods Instrum Comput, 1985, 17(5): 519~525.
[4] Martin PH and Unwin DM. A microwave Doppler radar activity monitor. Behav Res Methods Instrum, 1980, 12(5):517~520.
[5] Giansanti D, Maccioni G, Macellari V. The development and test of a device for the reconstruction of 3-D position and orientation by means of a kinematic sensor assembly with rate gyroscopes and accelerometers. IEEE Trans Biomed Eng, 2005, 52(3): 1271~1277.
[6] Georgopoulos AP, Schwartz AB, Kettner RE. Neuronal population coding of movement direction. Science, 1986, 233(4771): 1416~1419.
[7] M. Furniss. Motion capture. MIT Communication Forum, 1999.
[8] C. Grow, I. Gordon, R.D. Stuart, A. Adalja. Motion Capture as a Means for Data Acquisition .ARA Focus Group Spring, 1998.
[9] Macmillan DL. A physiological analysis of walking in the American lobster (Homarus americanus), Philos Trans R Soc Lond B Biol Sci , 1975, 270(901): 1~59.
[10] Uter TG. A real-time video system for tracking one- dimensional movements of two objects. IEEE Trans Biomed Eng, 1977, 24(1): 8~75.
[11] Pascual J.Figueroa , Neucimar J. Leite, and Ricardo M.L.Barros. A ?exible software for tracking of markers used in human motion analysis. Computer Methods and Programs in Biomedicine, 2003, 72(2): 155~165.
[12]程建,王东泉,施鹏飞等.基于图象技术的三维步态运动测量.上海交通大学学报, 1998, 32(4): 59~64.
[13]胡琴,王文中,夏时洪等.基于多摄像机的人体步态跟踪方法.计算机工程, 2008, 34(22): 220~222.
[14]王人成,黄昌华,王季军等.基于普通摄像机的人体运动信息检测系统.生物医学工程学杂志, 1999, 16 (4) : 448~452.
[15]王人成,董华,黄昌华等.低成本人体步态在线检测系统.清华大学学报, 2002, 42(2): 165~167.
[16]蔡付文,王人成,李广庆等.低成本人体步态分析系统的研制.康复医学工程, 2008, 23(1): 49~53.
[17] Haritaoglu I,Harwood D, Davis L Y. Real-time surveillance of people and their activities. IEEE Trans Pattern Analysis and Machine Intelligence, 2000, 22(8): 809~830 .
[18]李建华,孙怡,马小妹等.基于距离图像的人体下肢关节点运动跟踪.全国信号处理学术年会论文集, 2001: 333~336.
[19]冯莲,邹北骥,刘相滨等.无标志的人体步行腿部骨架检测与跟踪.计算机工程与科学, 2007, 29(1): 62~65.
[20]潘泉,梁彦,杨峰等.现代目标跟踪与信息融合.北京:国防工业出版社, 2009. 32~39
[21]彭亮.机动目标跟踪算法的研究:[硕士学位论文].华中科技大学图书馆, 2007.
[22]周宏仁,敬忠良,王培德.机动目标跟踪.北京:国防工业出版社, 1991. 10~142
[23] Moose R L. An adaptive state estimation solution to the maneuvering target problem. IEEE Trans. Autom. Control, 1975, AC-20(6): 359~362.
[24] Morgan D R. A target trajectory noise model for Kalman trackers. IEEE Transaction on Aerospace and Electronic Systems, 1976, 12(3): 405~408.
[25]杨耿,和卫星.运动目标图像识别与跟踪系统的研究.计算机测量与控制, 2005, 13(3): 267-269.
[26] Greg Welch, Gary Bishop. An Introduction to the Kalman Filter. SIGGRAPH, USA , 2001, Course 8:20~24.
[27]李尊民.电视图像自动跟踪的基本原理.北京:国防工业出版社, 1998. 41~74
[28]张庆生,邓兵,汪遵懋.低空高射武器观瞄系统的图像处理技术.光学技术,2001,27(5):262~265.
[29]刘宝生,闫莉萍,周东华.几种种经典相似性度量的比较研究.计算机应用研究, 2006, 11: 1~3.
[30]恽晓平.康复疗法评定学.北京:华夏出版社,2005.263~287
[31] Karim Fouad, Gerlinde .S. Metz, Doron Merkler etal. Treadmill training in incomplete spinal cord injured rats. Behavioural Brain Research,2000, 115(1):107~113.
[32] K.G. Pearson, H. Acharya, K. Fouad. A new electrode configuration for recordingelectromyographic activity in behaving mice. Journal of Neuroscience Methods, 2005, 148(1):36~42.
[33] A.J. Spink, R.A.J. Tegelenbosch, M.O.S. Buma,et al.The EthoVision video tracking system—A tool for behavioral phenotyping of transgenic mice.Physiology & Behavior,2001, 73(5):731~744.
[34] Chad V. Anderson, Andrew J. Fuglevand. Probability-Based Prediction of Activity in Multiple Arm Muscles:Implications for Functional Electrical Stimulation. Journal of Neurophysiol, 2008, 100(1): 482~494.
[35]贺兴华,周媛媛,王继阳等. MATLAB 7.X图像处理.北京:人民邮电出版社, 2006. 2~9
[36]杨青智,刘明治.机动目标跟踪技术的研究:[硕士学位论文].华中科技大学图书馆,2008.
[37]方青,梅晓春,张育平.用于机动目标跟踪的Kalman滤波器的设计.雷达科学与技术, 2006, 2(1): 50-55.
[38] Blair W D.Rice T R, Mcdole B S. Least-squares approach to asynchronous data fusion. SPIE Acquisition, Tracking and Pointing VI, USA, 1992, 1697(6):130~139.
[39]章毓晋.图像工程(上册).北京:清华大学出版社, 2001, 82~90.
[40]飞思科技产品研发中心. MATLAB 6.5辅助图像处理.北京:电子工业出版社, 2003. 59~69.
[39] Ian D. Peikona, Nathan A. Fitzsimmonsb, Mikhail A. Lebedev etal. Three-dimensional, automated,real-time video system for tracking limb motion in brain-machine interface studies. Journal of Neuroscience Methods. 2009, 180(2): 224~233.
[42] Gerald L. Scheirman and Phillip J. Cheetham. Motion measurement using the Peak Performance Technologies system. SPIE Image-Based Motion Measurement, 1990, 1356(11): 63~70.
[43] Yury P. Gerasimenko, Ronaldo M. Ichiyama,Igor A etal. Lavrov,Epidural Spinal Cord Stimulation Plus Quipazine Administration Enable Stepping in Complete Spinal Adult Rats. J Neurophysiol, 2007, 98(5): 2525~2536.