视频序列中运动目标的实时分割与跟踪
|
摘要
运动目标的分割与跟踪是当前计算机视觉中非常活跃的一个研究领域。计算机视觉研究的目标是使计算机具有通过一幅或多幅图像认知周围环境信息的能力。人类通过感官感知外界的信息，而视觉信息是人类获取的主要信息。运动图像序列中包含了比静态图像更多的有用信息。运动对象的分割，是将视频图像序列划分成若干的运动对象，并在时间轴上对这些运动对象进行跟踪，为以后的研究工作奠定基础，如基于对象的编码技术以及基于内容的视频检索等。目标跟踪则是随着数字视频技术的发展及应用而产生的一个新的研究课题，其在军事以及民用等诸多领域中有广泛的应用。
本论文研究的目的有两点：一是研究如何从视频流中正确的分割出所感兴趣的运动对象，二是研究如何对视频流中的人脸进行实时跟踪。
本论文首先研究了在静态背景中，对视频流中运动目标进行实时分割，通过帧间变化检测来检测运动目标对象，并分割出运动目标。在分割运动目标算法中，本课题实现了两种算法，并且同文献中的两种算法进行比较，最后分析试验结果。后处理采用了形态学算子，其方法简单有效，结果令人满意。
运动目标实时跟踪的研究是本论文的重点。在基于静态图像中使用的MSA(Mean Shift Algorithm)算法的基础上，本论文提出了DMSA(Dynamical Mean Shift Algorithm)跟踪算法，并结合肤色概率分布，将其应用于人脸跟踪的研究中。
西安理工大学硕士学位论文
最后在不同背景的视频流中测试了该算法的实时性、准确性等指标，实验结果令人
满意。
Robust segmentation and tracking is an active field of computer vision. The aid of computer vision make computer have the ability of learning about surrounding information. There is more useful information in motion sequences than static images. Segmentation of moving objects is to find differential objects in motion sequences and track them. This study is important, because it is the base of video retrieving based on content or video coding based on objects and so on. Object tracking is a new research topic with the developing of digital video technology. It has been generally used in military and civil.
In this paper there are two studying points: one is to study how to segment objects in video sequences, the other is how to make real-time objects tracking.
We can find out motion objects by changing detection in static background scenes. In this paper I bring forward two methods for segmentation and compare them with two traditional methods. At last, I use morphological approach to improve segmental quality.
Real-time tracking of moving object is the emphasis of this paper. Based on Mean Shift algorithm I propose Dynamical Mean Shift algorithm. In this paper,
I use DMSA to track face in video sequences which combines a probability distribution image of flesh color. In the end of this paper, DMSA's tracking accuracy, tolerance to noise, distractors and performance are studied.
本论文研究的目的有两点：一是研究如何从视频流中正确的分割出所感兴趣的运动对象，二是研究如何对视频流中的人脸进行实时跟踪。
本论文首先研究了在静态背景中，对视频流中运动目标进行实时分割，通过帧间变化检测来检测运动目标对象，并分割出运动目标。在分割运动目标算法中，本课题实现了两种算法，并且同文献中的两种算法进行比较，最后分析试验结果。后处理采用了形态学算子，其方法简单有效，结果令人满意。
运动目标实时跟踪的研究是本论文的重点。在基于静态图像中使用的MSA(Mean Shift Algorithm)算法的基础上，本论文提出了DMSA(Dynamical Mean Shift Algorithm)跟踪算法，并结合肤色概率分布，将其应用于人脸跟踪的研究中。
西安理工大学硕士学位论文
最后在不同背景的视频流中测试了该算法的实时性、准确性等指标，实验结果令人
满意。
Robust segmentation and tracking is an active field of computer vision. The aid of computer vision make computer have the ability of learning about surrounding information. There is more useful information in motion sequences than static images. Segmentation of moving objects is to find differential objects in motion sequences and track them. This study is important, because it is the base of video retrieving based on content or video coding based on objects and so on. Object tracking is a new research topic with the developing of digital video technology. It has been generally used in military and civil.
In this paper there are two studying points: one is to study how to segment objects in video sequences, the other is how to make real-time objects tracking.
We can find out motion objects by changing detection in static background scenes. In this paper I bring forward two methods for segmentation and compare them with two traditional methods. At last, I use morphological approach to improve segmental quality.
Real-time tracking of moving object is the emphasis of this paper. Based on Mean Shift algorithm I propose Dynamical Mean Shift algorithm. In this paper,
I use DMSA to track face in video sequences which combines a probability distribution image of flesh color. In the end of this paper, DMSA's tracking accuracy, tolerance to noise, distractors and performance are studied.
引文
[1] 章毓晋．图象处理和分析．清华大学出版社．2001．
[2] 马颂德，张正友．计算机视觉—计算理论预算法基础．科学出版社 1998
[3] 季百杨 陈纯 钱英·视频分割技术的发展·计算机研究与发展 2001．1
[4] 韩军 熊璋 李超 龚声蓉·分割视频运动对象的研究·计算机工程与应用 2000，8
[5] Jean-Yves Bouguet. Pyramidal Implementation of the Lucas Kanade Feature Tracker Description of the algorithm. Intel Corporation, Microprocessor Research Labs. 1998
[6] 刘李杰 蔡德钧 翁南钐·一种面向运动的视频对象分割算法·计算机学报 2000．12
[7] 史立 张兆扬·面向视频编码的运动对象分割和提取·上海大学学报 Vol．7．No．1 2001．2
[8] Alan J. Lipton Hironobu. Fujiyoshi Raju S. Patil. Moving Target Classification and from Real-time Video. The Robotics Institute Carnegie Mellon University 2000
[9] 黄友珍 黄艺 余兆明·基与修正分水岭算法和时域跟踪的视频自动分割·数字视频 2000．1
[10] 李海明 陈新 吴芳 赵音频·复杂背景下运动目标的光流区域提取方法·福州大学学报 2001．8
[11] 俞毅刚 龚建荣·基与多个非刚体目标跟踪的视频对象平面生成算法·数字视频 2001．1
[12] Bala L P, Talml K, Liu Jj. Automatic detection and tracking of faces and facial teatures in video sequences. Picture Coding Symposium 1997.1997, 251～256
[13] Antozyn P M, Hannah J M, Grant P M. Tracking of the motion of important facial fetures in model-based coding. Signal Processing, 1998:249～260
[14] 潘金辉 廖庆敏 林行刚·视频序列中运动目标的自动提取·清华大学学报 2001．5
[15] A. Selinger, L. Wixson. Classifying moving objects as rigid or non-rigid without Correspondences. 2001.
[16] Shu-Ching Chen, Mei-Ling Shyu. Video Scene change Detection Method using Unsupervised Segmentation and Objet Tracking. IEEE. 2001
[17] 汪力新，戴汝为·三维仿射不变距·模式识别与人工智能 1998：133～139
[18] 崔屹等·数学形态学方法及应用·科学出版社 2000
[19] 王拴，艾海舟，和克忠·基于差分图像的多运动目标的检测与跟踪·中国图象图形学报 1999．6
[20] 周长发·精通VC图象处理编程·电子工业出版社 2000
[21] 周锐，杨涤·复杂背景中目标图像的提取与跟踪·系统工程与电子技术 1997：10～13
[22] 夏玮，李朝晖·中值滤波的快速算法·计算机工程与设计 Vol．23 No．1 2002．1：58～59
[23] 何东建，庚楠·中值滤波的快速算法的探讨与实验·微电脑应用 No．3 1998：32～34
[24] Sohaib Khan, Mubarak Shah. Object Based Segmentation of Video Using Color, Motion and Spatial Information. Computer Vision Laboratory 2001
[25] 刘忠伟，章毓晋·十种基于颜色特征图像检索算法的比较和分析·信号处理 Vol，16．No．1 2000．3：79～84
[26] 张毅军，吴雪箐，夏良正·二维熵图像阈值分割的快速递推算法·模式识别与人工智能 Vol．10．No．3 1997：259～264
[27] Loannis Kompatsiaris. Spatiotemporal Segmentation and Tracking of Objects for Visualization of Videoconference Image Sequences. IEEE 2000
[28] J urgen Stauderl, Roland Mech. Criteria for Image Regions Changed by Moving Shadows. IRISA/INRIA. 2001
[29] Sheng-Jyh Wang. The Study of Image Segmentation Using Color Information. 2000.10
[30] P. Salembier. Region-based representations of image and video: Segmentation tools for multimedia services. IEEE. 2000
[31] Berkeley. Video Motion Capture. UCB//CSD.2000
[32] 王亮，胡卫明，谭铁牛·人运动的视觉分析综述·《计算机学报》2002
[33] 范勇，游志胜，张建州，郑文琛，冯子亮·一种快速运动目标检测与跟踪算法·光电工程 Vol．27．No．6 2000．12：30～33
[34] 牛朝玮，汪增福·基于彩色和运动信息的人脸检测·模式识别与人工智能 Vol．15．No．2 2002．6：205～210
[35] 张星明·视频图像捕捉及运动检测技术的实现·计算机工程 Vol．28．No．8 2002．8：130～132
[36] 潘金辉 廖庆敏 林行刚·视频序列中运动目标的自动提取·清华大学学报 2001．5：190～193
[37] 卢官明·区域生长型分水岭算法及其在图像序列分割中的应用·南京邮电学院学报 Vol．20．No．3 2000．9：51～54
[38] Apostolos Dailianas. Comparison of Automatic Video Segmentation Algorithms. IEEE 1998
[39] Chenyang Xu, Jerry L. Prince. Gradient Vector Flow:A NewExternal Force for Snakes. IEEE Proc. Conf. Vis. Patt. Recog. 1997
[40] J. Denzler, H. Niemann. Evaluating the Performance Of Active Contour Models for Real-time Object Tracking. 1996
[41] Hager G D, Belhumeur P N. Efficient region tracking with parametric models of geometry and illumination. IEEE Trans. PAMI. 1998, 20:1025～1039
[42] Kervrann C, Heitz F.a hierarchical Markov modeling approach for the
segment and tracking of deformable shapes. Grahical Models And Image Processing, 1998:173～195
[43] 高文，王瑞，马继涌·一种快速、鲁棒的唇动检测与定位方法·计算机学报 2001．8
[44] R.L. Hsu, A. M. Mohamed and A. K. Jain. Face Detection in Color Images, ICIP'2001
[45] 梁路宏，艾海舟，徐光祜，张钹·基于肤色和模板的人脸检测·软件学报 2001．12
[46] 陶霖密，彭振云，徐光祜·人体的肤色特征·软件学报 2000
[47] 沈兰荪，卓力，田栋，汪孔桥·视频编码与低速率传输·电子工业出版社 2001
[48] D. Comaniciu, P. Meer. Mean Shift Analysis and Applications. IEEE int' 1 Conf. Comp. 1999
[49] Peter Meer. Real-Time Tracking of Non-Rigid Object using Mean Shift. IEEE Conf. on Computer Vision and Pattern Recognition. 2000
[50] Dorin Comaniciu. Mean Shift:A Robust Approach Toward Feature Space Analysis. IEEE Trans. Pattern Anal. Machine Intell. vol. 24, no. 5, pp. 603-619, 2002.5
[51] S. Birchfield. Elliptical Head Tracking using intensity Gradients and Color Histograms. IEEE Conf. on Computer Vision and Pattern Recognition. 1998
[52] Ferran Manques, Cristina Molina. Object Tracking for Content-Based Functionalities. 2000
[53] 刘珂 张宪民 付永会·一种改进的Hausdorff距离目标跟踪算法．上海交通大学学报 2001．2
[54] 李海明 陈新 吴芳 赵音频·复杂背景下运动目标的光流区域提取方法·福州大学学报 2001．8
[55] 关海英 阮秋琦·基于Hausdorff距离的非刚体目标自适应轮廓跟踪·通信学报 1998．11
[56] 詹翊强 戚飞虎 刘天明·基于HPEG—2视频流的目标跟踪快速算法·上海交通大学学报 2001．9
[57] Fabrice Moscheni, Frederic Dufaux, Murat Kunt. Object Tracking Based on Temporal and Spatial Information. Signal Processing Laboratory 1996.
[58] Hogg D·Model-based vision:A program to see a walking persion·Image and Vision Computing·1995
[59] Simon W L, "A Multiple Target Tracking System", SPIE Vol. 1388 Mobile Robots V, 1990:pp299-305
[60] N. Paragios, G. Tziritas. Detection and Location of Moving Objects using deterministic relaxation algorithms. IEEE. 1997
[61] Dorin Comaniciu, Visvanathan Ramesh, Peter Ueer. Kernel-Based Object Tracking. IEEE Conference on Computer Vision and Pattern Recognition,. 2003
[62] B. Bascle and R. Deriche, "Region tracking through image sequences," in Proc. 5th Intl. Conf. on Computer Vision, Cambridge, MA, 1995:302-307.
[63] Isaac Cohen, G'erard medioni. Detecting and Tracking Moving Objects for Video Surveillance. IEEE Proc. Computer Vision and Pattern Recognition. 1999
[64] Walter W. Bell, Pedro F. Felzenszwalb and Daniel P. Huttenlocher. Detection and Long Term Tracking of Moving Objects in Aerial Video. 1999
[65] Andrea Prati, Ivana Mikillc, Costantino Grana, Mohan M. Trivedi. Shadow Detection Algorithms for Tracffic Flow Analysis: a Comparative Study. 2001
[66] Dorin Comaniciu. Bayesiam Kernel Tracking. IEEE Proc. Computer Vision and Pattern Recognition. 2002
[67] Dorin Comaniciu and Visvanathan Ramesh. Mean Shift and Optimal Prediction For Efficient Object Tracking. IEEE. 2000
[68] 陈震，高满屯，沈允文·图象光流场计算技术研究进展·中国图象图形学报 Vol．7 No．5 2002
[69] 刘江华，程君实，程佳晶·基于光流的动态手势识别·计算机工程 Vol．28 No．4 2002
[2] 马颂德，张正友．计算机视觉—计算理论预算法基础．科学出版社 1998
[3] 季百杨 陈纯 钱英·视频分割技术的发展·计算机研究与发展 2001．1
[4] 韩军 熊璋 李超 龚声蓉·分割视频运动对象的研究·计算机工程与应用 2000，8
[5] Jean-Yves Bouguet. Pyramidal Implementation of the Lucas Kanade Feature Tracker Description of the algorithm. Intel Corporation, Microprocessor Research Labs. 1998
[6] 刘李杰 蔡德钧 翁南钐·一种面向运动的视频对象分割算法·计算机学报 2000．12
[7] 史立 张兆扬·面向视频编码的运动对象分割和提取·上海大学学报 Vol．7．No．1 2001．2
[8] Alan J. Lipton Hironobu. Fujiyoshi Raju S. Patil. Moving Target Classification and from Real-time Video. The Robotics Institute Carnegie Mellon University 2000
[9] 黄友珍 黄艺 余兆明·基与修正分水岭算法和时域跟踪的视频自动分割·数字视频 2000．1
[10] 李海明 陈新 吴芳 赵音频·复杂背景下运动目标的光流区域提取方法·福州大学学报 2001．8
[11] 俞毅刚 龚建荣·基与多个非刚体目标跟踪的视频对象平面生成算法·数字视频 2001．1
[12] Bala L P, Talml K, Liu Jj. Automatic detection and tracking of faces and facial teatures in video sequences. Picture Coding Symposium 1997.1997, 251～256
[13] Antozyn P M, Hannah J M, Grant P M. Tracking of the motion of important facial fetures in model-based coding. Signal Processing, 1998:249～260
[14] 潘金辉 廖庆敏 林行刚·视频序列中运动目标的自动提取·清华大学学报 2001．5
[15] A. Selinger, L. Wixson. Classifying moving objects as rigid or non-rigid without Correspondences. 2001.
[16] Shu-Ching Chen, Mei-Ling Shyu. Video Scene change Detection Method using Unsupervised Segmentation and Objet Tracking. IEEE. 2001
[17] 汪力新，戴汝为·三维仿射不变距·模式识别与人工智能 1998：133～139
[18] 崔屹等·数学形态学方法及应用·科学出版社 2000
[19] 王拴，艾海舟，和克忠·基于差分图像的多运动目标的检测与跟踪·中国图象图形学报 1999．6
[20] 周长发·精通VC图象处理编程·电子工业出版社 2000
[21] 周锐，杨涤·复杂背景中目标图像的提取与跟踪·系统工程与电子技术 1997：10～13
[22] 夏玮，李朝晖·中值滤波的快速算法·计算机工程与设计 Vol．23 No．1 2002．1：58～59
[23] 何东建，庚楠·中值滤波的快速算法的探讨与实验·微电脑应用 No．3 1998：32～34
[24] Sohaib Khan, Mubarak Shah. Object Based Segmentation of Video Using Color, Motion and Spatial Information. Computer Vision Laboratory 2001
[25] 刘忠伟，章毓晋·十种基于颜色特征图像检索算法的比较和分析·信号处理 Vol，16．No．1 2000．3：79～84
[26] 张毅军，吴雪箐，夏良正·二维熵图像阈值分割的快速递推算法·模式识别与人工智能 Vol．10．No．3 1997：259～264
[27] Loannis Kompatsiaris. Spatiotemporal Segmentation and Tracking of Objects for Visualization of Videoconference Image Sequences. IEEE 2000
[28] J urgen Stauderl, Roland Mech. Criteria for Image Regions Changed by Moving Shadows. IRISA/INRIA. 2001
[29] Sheng-Jyh Wang. The Study of Image Segmentation Using Color Information. 2000.10
[30] P. Salembier. Region-based representations of image and video: Segmentation tools for multimedia services. IEEE. 2000
[31] Berkeley. Video Motion Capture. UCB//CSD.2000
[32] 王亮，胡卫明，谭铁牛·人运动的视觉分析综述·《计算机学报》2002
[33] 范勇，游志胜，张建州，郑文琛，冯子亮·一种快速运动目标检测与跟踪算法·光电工程 Vol．27．No．6 2000．12：30～33
[34] 牛朝玮，汪增福·基于彩色和运动信息的人脸检测·模式识别与人工智能 Vol．15．No．2 2002．6：205～210
[35] 张星明·视频图像捕捉及运动检测技术的实现·计算机工程 Vol．28．No．8 2002．8：130～132
[36] 潘金辉 廖庆敏 林行刚·视频序列中运动目标的自动提取·清华大学学报 2001．5：190～193
[37] 卢官明·区域生长型分水岭算法及其在图像序列分割中的应用·南京邮电学院学报 Vol．20．No．3 2000．9：51～54
[38] Apostolos Dailianas. Comparison of Automatic Video Segmentation Algorithms. IEEE 1998
[39] Chenyang Xu, Jerry L. Prince. Gradient Vector Flow:A NewExternal Force for Snakes. IEEE Proc. Conf. Vis. Patt. Recog. 1997
[40] J. Denzler, H. Niemann. Evaluating the Performance Of Active Contour Models for Real-time Object Tracking. 1996
[41] Hager G D, Belhumeur P N. Efficient region tracking with parametric models of geometry and illumination. IEEE Trans. PAMI. 1998, 20:1025～1039
[42] Kervrann C, Heitz F.a hierarchical Markov modeling approach for the
segment and tracking of deformable shapes. Grahical Models And Image Processing, 1998:173～195
[43] 高文，王瑞，马继涌·一种快速、鲁棒的唇动检测与定位方法·计算机学报 2001．8
[44] R.L. Hsu, A. M. Mohamed and A. K. Jain. Face Detection in Color Images, ICIP'2001
[45] 梁路宏，艾海舟，徐光祜，张钹·基于肤色和模板的人脸检测·软件学报 2001．12
[46] 陶霖密，彭振云，徐光祜·人体的肤色特征·软件学报 2000
[47] 沈兰荪，卓力，田栋，汪孔桥·视频编码与低速率传输·电子工业出版社 2001
[48] D. Comaniciu, P. Meer. Mean Shift Analysis and Applications. IEEE int' 1 Conf. Comp. 1999
[49] Peter Meer. Real-Time Tracking of Non-Rigid Object using Mean Shift. IEEE Conf. on Computer Vision and Pattern Recognition. 2000
[50] Dorin Comaniciu. Mean Shift:A Robust Approach Toward Feature Space Analysis. IEEE Trans. Pattern Anal. Machine Intell. vol. 24, no. 5, pp. 603-619, 2002.5
[51] S. Birchfield. Elliptical Head Tracking using intensity Gradients and Color Histograms. IEEE Conf. on Computer Vision and Pattern Recognition. 1998
[52] Ferran Manques, Cristina Molina. Object Tracking for Content-Based Functionalities. 2000
[53] 刘珂 张宪民 付永会·一种改进的Hausdorff距离目标跟踪算法．上海交通大学学报 2001．2
[54] 李海明 陈新 吴芳 赵音频·复杂背景下运动目标的光流区域提取方法·福州大学学报 2001．8
[55] 关海英 阮秋琦·基于Hausdorff距离的非刚体目标自适应轮廓跟踪·通信学报 1998．11
[56] 詹翊强 戚飞虎 刘天明·基于HPEG—2视频流的目标跟踪快速算法·上海交通大学学报 2001．9
[57] Fabrice Moscheni, Frederic Dufaux, Murat Kunt. Object Tracking Based on Temporal and Spatial Information. Signal Processing Laboratory 1996.
[58] Hogg D·Model-based vision:A program to see a walking persion·Image and Vision Computing·1995
[59] Simon W L, "A Multiple Target Tracking System", SPIE Vol. 1388 Mobile Robots V, 1990:pp299-305
[60] N. Paragios, G. Tziritas. Detection and Location of Moving Objects using deterministic relaxation algorithms. IEEE. 1997
[61] Dorin Comaniciu, Visvanathan Ramesh, Peter Ueer. Kernel-Based Object Tracking. IEEE Conference on Computer Vision and Pattern Recognition,. 2003
[62] B. Bascle and R. Deriche, "Region tracking through image sequences," in Proc. 5th Intl. Conf. on Computer Vision, Cambridge, MA, 1995:302-307.
[63] Isaac Cohen, G'erard medioni. Detecting and Tracking Moving Objects for Video Surveillance. IEEE Proc. Computer Vision and Pattern Recognition. 1999
[64] Walter W. Bell, Pedro F. Felzenszwalb and Daniel P. Huttenlocher. Detection and Long Term Tracking of Moving Objects in Aerial Video. 1999
[65] Andrea Prati, Ivana Mikillc, Costantino Grana, Mohan M. Trivedi. Shadow Detection Algorithms for Tracffic Flow Analysis: a Comparative Study. 2001
[66] Dorin Comaniciu. Bayesiam Kernel Tracking. IEEE Proc. Computer Vision and Pattern Recognition. 2002
[67] Dorin Comaniciu and Visvanathan Ramesh. Mean Shift and Optimal Prediction For Efficient Object Tracking. IEEE. 2000
[68] 陈震，高满屯，沈允文·图象光流场计算技术研究进展·中国图象图形学报 Vol．7 No．5 2002
[69] 刘江华，程君实，程佳晶·基于光流的动态手势识别·计算机工程 Vol．28 No．4 2002