用户名: 密码: 验证码:
运动目标跟踪算法的研究
详细信息    本馆镜像全文|  推荐本文 |  |   获取CNKI官网全文
摘要
运动目标的自动跟踪技术是一项融合图象处理、模式识别、人工智能、自动控制等多种不同领域先进成果的高技术课题,是实现智能机器人和智能化武器等的关键技术之一,在军事、交通、生物医学等多种领域都有广泛应用。
     本文通过大量的实验主要研究了3种跟踪方法:(1)差分法;(2)相关法;(3)主动轮廓线。
     本文先介绍了一种比较简单的跟踪方法——差分法,即将两幅图象做“相减”运算,从相减后的图象中,得到运动物体的信息。这种方法操作起来比较简单,直接对两幅图像做差,然后再阈值分割即可。但采用这种方法的前提条件是须保证背景绝对静止或基本无变化(噪声较小),而在实际的运动图像中,光照的变化、噪声等因素的干扰是不可避免的,致使运动目标的检测和跟踪变得很不可靠。
     基于相关的跟踪方法,也叫相关匹配法或模板匹配法。对于这种方法,先研究了普通匹配,由于其匹配过程非常耗时,根本无法用于实时跟踪,故紧接着又研究了速度较快的多分辨率匹配,即先在分辨率较低的图像上粗匹配,得到一些侯选区域,然后在这些侯选区域内再进行精匹配,得到最终的匹配结果,实现对运动目标的跟踪。针对多分辨率的方法,又分成跳跃式多分辨率(Jump Multi-resolution:JMR)和平滑式多分辨率(Smooth Multi-resolution:SMR)分别进行研究。由于上面这两种方法的模板图像固定,因而对目标运动状态变化的自适应性非常差,故下文又介绍了一种加入“更新”和“预测”机制的匹配跟踪方法——结合Kalman滤波器的匹配方法,使跟踪更准、更快。所谓“更新”,就是对模板图象不断更新,把当前帧的匹配区域作为新的模板,在下一帧中用该更新后的模板进行匹配,再把匹配区域作为新的模板,如此反复,模板就能适应目标状态的不断变化,所以跟踪“更准”;所谓“预测”,就是充分利用动态图象序列之间的相关性,根据目标以前的数据(位置、速度)估计目标将来可能的位置区域,在该区域里进行多分辨率匹配,这样就减少了目标的匹配区域,从而缩短了匹配时间,使跟踪“更快”。
    
    沈阳l一业人学硕十学位论文
     近十几年来,主动轮廓线(Snake)J卜始被应用到目标跟踪中,很多研究表明它很
    适合于刚体和非刚体的跟踪。木文先从主动轮廓线的定义和计算入手,分析了3种不同
    的计算方法,最后选定较好的贪婪算法。然后以此为基础,通过蛇点的重抽样,气球
     (Balloon)模型,各能量项对s几永e的影响,噪声对s斑永e的影响等不同侧面深入研究
    了它的运行机制,最后给出用它进行目标跟踪的实验结果。为了改进原始模型中蛇点固
    定不变的缺点,本文引入重抽样机制,自适应地增加或删减蛇点,使分辨率在snake的
    进化过程中始终保持恒定。接着,又讨论了Balloon模型,并针对其缺点引入改进的
    BaUoon模型。各能量项确定以后,本文对各能量项的系数对s班tke的影响又进行了较为
    深入的实验研究。在实际的图像采集系统中,噪声是不可避免的,为此,文章以椒盐噪
    声为例,研究了它们对snake的影响,并举出一些去除噪声的方法,如中值滤波和自适
    应平滑滤波等。最后,在__卜面研究的基础上,介绍了snake在本文中的两个应用一目标
    跟踪和虹膜定位。
Automatic tracking of moving target is high technological subject which employs advanced achievements in many fields such as image processing, pattern recognition, artificial intelligence and automatic control, etc. It has wide application in many fields such as intelligent automatic weapn, intelligent robots, autonomous vehicle guidance, biology and so on.
    In this paper, three tracking methods are researched with the help of many experiments: (1) difference, (2) correlation, (3) active contour.
    First, difference is studied. Its main idea is that two images are subtracted, and some information of moving target is derived from the difference image. Obviously, this method is very simple and convenient to operate. However, it also has its drawbacks. The premise of application of this method is that background should be quite stable and has no great changes. But in practical images, the illumination changes and the noise from around environment are unavoidable, which result in low reliability when detecting and tracking moving target.
    Second, correlation-based method, also called template-matching method, is explored. For one thing, ordinary matching without any improved methods is researched, and it is found that this method cannot be applied to tracking moving target at all due to huge time consumption. Therefore, multi-resolution matching is researched. Its main idea is that coarse matching is operated in the lower resolution image, obtaining some waiting regions, then fine matching is applied to the waiting images with higher resolution, and find the position of moving target. JMR (Jump Multi-Resolution) and SMR (Smooth Muhi-Resolution) are studied respectively. Because the templates of above matching methods are both changeless, they have poor self-adaptability to the changes of moving target. Accordingly, an improved method integrated with Kalman Filter with better performance is introduced, which has two key techniques-prediction and update. Prediction's main idea is to take full advantage of correlation between frames, and to pred
    ict the possible positions of moving target based on previous data. The matching is just
    
    
    completed in the possible region, which improve its speed due to reduction of waiting region. Update's main idea is that the matched region in current frame is regarded as the template of moving target for next frame and repeats like this, which makes the template updated constantly and has better adaptability to the change of target.
    In recent ten years, active contour (also called "snake") has begun to be applied to moving target tracking. Investigation shows that "snake" is suitable for rigid or nonrigid target tracking. Starting with researching on defining and computing of snake, three computing method are analyzed, and Greedy Algorithm is chosen as the final method. Based on above, the operation mechanism of snake is explored deeply from different view such as resampling of snake pixel (snaxel), Balloon model, each energy's influence on snake, noise's influence on snake and so oa In order to avoid the drawback of invarible snaxel in original model, snaxel resampling is introduced, which can increase or reduce the number of snaxel adaptively and render the resolution of snake keep invarible during its evolution. Followed that, Ballon model and its improved model are studied successively. After determination of each energy item, their weights' influences on snake are researched through a great deal of experiments. In pratical image-ca
    ptureing system, noise is inevitable; therefore, salt and pepper noises' influences on snake are researched respectively, and several methods of noise reducing like median value filtering and self-adaptive smooth filtering are introduced. Finally, two applications of snake-tracking of moving target and location of iris-are presented.
引文
[1] 杨杨。自然景物环境下运动目标的自动跟踪方法。哈尔滨:哈尔滨工业大学博士学位论文,1998。
    [2] 王国营,张红旗等。一种成像跟踪系统的设计。计算机工程,2000,26(2):32~33。
    [3] 王春平,朱元昌等。基于图象信息的跟踪算法分析。火力与指挥控制,2000,25(1):19~21。
    [4] 熊国清,于起峰。用于实时跟踪的快速匹配算法。计算机辅助设计与图形学报,2002,14(1):41~43。
    [5] 李强,杨涛。形心和匹配跟踪算法的改进。光电工程,2000,27(2):28~30。
    [6] 刘铁军,肖颖杰等。一种改进的区域相关跟踪方法。山东科技大学学报(自然科学版),2001,20(3):44~48。
    [7] 任仙怡,周晓等。基于仿射变换与模板更新策略的目标跟踪方法。华中科技大学学报,2001,29(12):50-51。
    [8] 赖宏彬,黄卫农等。基于静止分割和运动矢量的对象跟踪。计算机工程,2001,27(7):140~142。
    [9] 詹翊强,臧飞虎等。基于MPEG-2视频流的目标跟踪快速算法。上海交通大学学报,2001,35(9):1321~1324。
    [10] 付永会,张风超等。一种改进的基于颜色直方图的实时目标跟踪算法。数据采集与处理,2001,16(3):309~314。
    [11] 郑立。自动跟踪算法的研究。武警工程学院学报,2002,18(2):39~42。
    [12] 杨昀,张桂林。一种新的相关跟踪算法的设计与实现。红外与毫米波学报,2000,19(5):377~380。
    [13] 章毓晋。图象分割。北京:科学出版社,2001。
    [14] 胡明昊。复杂场景下运动目标的实时跟踪。南京:南京理工大学硕士学位论文,2002。
    
    
    [15] 汪亚明,楼正国,卞昕,汪元美。一种非刚体运动图象序列的特征点对应方法。中国图象图形学报,2000,5(3):232~236。
    [16] 张岩,崔智杜等。图象序列中机动日标的形心跟踪。航空学报,2001,22(4):312~316。
    [17] 张嫣,姚耀文等。运动图像序列中多月标跟踪的研究与实现。计算机应用研究,2002,(1):74~76。
    [18] 陈朝阳,丁明跃等。基于对数极坐标映射的目标跟踪方法。电子学报,2001,29(9):1173~1175。
    [19] 栾新,朱铁一。快速搜索任意形状二维目标质心策略。中国图象图形学报,1999,4(5):372~376。
    [20] 韩艳丽,郭晓军等。可视化目标特征提取及动态跟踪算法验证系统。红外与激光工程,2001,30(2):99~101。
    [21] 杜亚娟,潘泉等。一种新的不变矩特征在图象识别中的应用。系统工程与电子技术,1999(10)。
    [22] 张坤华,王敬儒等。基于扩展目标的不变矩跟踪算法。强激光与粒子束,2002,14(1):6~10。
    [23] 任仙怡,周晓等。基于奇异值分解的图象目标跟踪算法。红外与激光工程,2001,30(6).438~441。
    [24] Shearer. K, Wong. K. D. Combining multiple tracking algorithms for improved general performance. Pattern Recognition, 2001,34:1257~1269.
    [25] Veenman.C.J, Reinders. M. J. T. Resolving Motion Correspondence for Densely Moving Points. IEEE Trans. Pattern Analysis and Machine Intelligence, 2001,23(1): 54~72.
    [26] 张安民,杨世兴等。目标跟踪中的混合多模方法-综述。系统工程与电子技术,2001,23(4):33~35。
    [27] 何衍,蒋静坪等。基于新息偏差的自适应机动目标跟踪算法。信息与控制,2001,30(4):333~336。
    
    
    [28] 范涛,杨晨阳等。基于模型混合的多分辨率多模型跟踪算法。北京航空航天大学学报,2001,27(1):28~31。
    [29] 徐毓,金以慧等。基于强跟踪滤波器的多目标跟踪方法。传感器技术,2002,21(3):17~20。
    [30] Tissainayagam.P, Suter.D. Visual tracking with automatic motion model switching. Pattern Recognition 34(2001): 641~660.
    [31] McKenna.S.J, Raja.Y. Tracking colour objects using adaptive mixture models. Image and Vision Computing, 17(1999): 225~231.
    [32] Mazor.E, Averbuch.A. Interacting multiple model methods in target tracking: a survey. IEEE Trans. On Aerospace and Electronic Systems, 1998, 34(1): 103~123.
    [33] Marques.J.S, Lemos.J.M. Optimal and suboptimal shape tracking based on multiple switched dynamic models. Image and Vision Computing, 2001,19: 539~550.
    [34] Toyama.K, Blake. A. Probabilistic Tracking with Exemplars in a Metric Space. International Journal of Computer Vision, 2002, 48(1): 9~19.
    [35] Vincze.M. Robust tracking of ellipses at frame rate. Pattern Recognition 34(2001): 487~498.
    [36] Kass.M, Witkin.A, and Terzopoulos.D. Snakes: Active Contour Models. The 1st International Conference of Computer Vision, 1987: 259~268.
    [37] Cohen. L.D. On Active Contour Models and Balloon. CVGIP: Image Understanding, 1991, 53 (2): 211~218.
    [38] Lobregt.S, Viergever.M.A. A Discrete Dynamic Contour Model. IEEE Trans. on Medical Imaging, 1995, 14(1): 12~24.
    [39] Xu.C, Prince.J.L. Snakes, shapes, and gradient vector flow. IEEE Trans. on Image Processing, 1998, 7: 359~369.
    [40] Xu.C, Prince.J.L. Generalized gradient vector flow external forces for active contours. Signal Processing, 1998, 71: 131~139.
    [41] Amini.A.A, Tehran.S, and Weymouth.T.E. Using Dynamic Programming for Minimizing the Energy of Active Contours in the Presence of Hard Constrains. The 2nd International Conference of Computer Vision, 1988: 95~99.
    
    
    [42] Amini.A.A, Weymouth.T.E, and Jain.R.C. Using Dynamic Programming for Solving Variational Problems in Vision. IEEE Trans.. on Pattern Analysis and Machine Intelligence, 1990, 12(9): 855~867.
    [43] Williams.D.J, Shah.M. A Fast Algorithm for Active Contours and Curvature Estimation. CVGIP: Image Understanding, 1992, 55(1): 14~26.
    [44] Terzopoulos.D, Waters.K. Analysis of Facial Images Using Physical Ans anatomical Models. Third Int. Conf. on Computer Vision. Osaka, Japan, 1990: 727~732.
    [45] Curwen.R. M, Blake.A and Cipolla.R. Parallel Implementation of Lagrangian Dynamics for Real-time Snakes. British Machine Vision Conference. England, 1991: 29~35.
    [46] Ray.N, Chanda.B. A fast and flexible multiresolution snake with a definite termination criterion. Pattern Recognition, 2001, 34: 1483~1490.
    [47] Kim.W, Lee.C.Y and Lee. J.J. Tracking moving object using Snake's jump based on image flow. Mechatronics, 2001, 11: 199~226.
    [48] Lilian Ji, Hong Yan. Attractable snakes based on the greedy algotirhm for contour extractiorL Pattem Recognition, 2002, 35: 791~806.
    [49] Jang.D.S, Choi.H.I1. Active models for tracking moving objects. Pattern Recognition, 2000(33), 1135~1146.
    [50] Pardas.M, Sayrol.E. Motion estimation based tracking of active contours. Pattern Recognition Letters, 2001,22: 1447~1456.
    [51] Fu.Y, Erdem.T. Tracking Visible Boundary of Objects Using Occlusion Adaptive Motion snake. IEEE Trans. On Image Processing, 2000, 9(12): 2051~2060.
    [52] Peterfreund.N. The velocity snake: Deformable contour tracking in spatio-velocity space. Computer Vision and Image Understanding, 1999,73(3): 346~356.
    [53] Zhong.Y, Jain, A.K, Object Tracking Using Deformable Template. IEEE Trans. On Pattern Analysis and Machine Intelligence, 2000,22(5): 544~549.
    [54] Pierre.M, Jolly.D. Tracking Deformable Templates Using a Shortest Path Algotithm. Computer Vision and Image Understanding, 2001 (81): 26~45.
    [55] Bertalmio.M, Sapiro.G. Morphing Active Contours. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2000, 22(7): 733~737.
    
    
    [56] Paragios.N, Deriche.R. Geodesic Active Contours and Level Sets for the Detection and Tracking of Moving Objects. IEEE Trans. On Pattern Analysis and Machine Intelligence, 2000, 22 (3): 266~280.
    [57] Delingette.H, Montagnat.J. Shape and Topology Constraints on Parametric Active Contours. Computer Vision and Image Understanding, 2001,83: 140~171.
    [58] Goldenberg.R, Kimmel.R. Fast Geodesic Active Contours. IEEE Trans. On Image Processing, 2001,10(10): 1467~1474.
    [59] Molloy.D, Whelan.P.F. Active-meshes. Pattern Recognition Letters, 2000(21): 1071~1080。
    [60] Ray.N, Chanda.B. A fast and flexible muttiresolution snake with a definite termination criterion Pattern Recognition, 2001(34): 1483~1490.
    [61] Delingette.H, Montagnat.J. Shape and Topology Constraints on Parametric Active Contours. Computer Vision and Image Understanding, 2001,83:140~171.
    [62] Pardas.M, Sayrol.E. Motion estimation based tracking of active contours. Pattern Recognition Letters, 2001,22: 1447~1456.
    [63] Lilian.J, Hong.Y. Attractable snakes based on the greedy algorithm for contour extraction. Pattern Recognition, 2002,35:791~806.
    [64] Iannizzotto.G, Vita.L. Real-time object tracking with movels and affine transformation. Image Processing, 2000. Proceedings. 2000 International Conference on: 316~318.
    [65] 孙晓阳。SNAKE——主动轮廓模型的算法研究与应用。上海:上海交通大学硕士学位论文,1997。
    [66] 刘志俭。主动轮廓模型的算法研究与应用。长沙:国防科学技术大学硕士学位论文,1999。
    [67] 张爱东,张田文。一种B-样条主动轮廓线模型。哈尔滨师范大学自然科学学报,1999,15(4):62~66。
    [68] 冯炯,戚飞虎。Snake成长阶段的分析方法。上海交通大学学报,1999,33(9):1149~1153。
    [69] 李熙莹,倪国强。一种自动提取目标的主动轮廓法。光子学报,2002,31(5):606~610。
    
    
    [70] 戚飞虎,沈定刚。基于边缘方向信息的主动轮廓算法。红外与毫米波学报,1997,16(1):45~50。
    [71] 杨杨,张田文。基于多分辨率方法的主动轮廓线跟踪算法。计算机学报,1998,21(3):210~216。
    [72] 吕铁英,彭嘉雄。图象轮廓特征提取新方法研究。中国图象图形学报,1999,4(8):655~658。
    [73] 赵雪春,戚飞虎。基于可变形模型及像素相似性统计的自动图像分割。红外与毫米波学报,2000,19(1):33~37。
    [74] 黄海赟,戚飞虎。基于多尺度图像的主动轮廓线模型。计算机研究与发展,2000,37(10):1240~1245。
    [75] 李培华,张阳文。主动轮廓线模型(蛇模型)综述。软件学报,2000,11(6):751~757。
    [76] 陈洪,王瀚,王燕燕等。作用域扩展函数在主动轮廓线模型中的应用。计算机工程,2002,28(8):86~87。
    [77] 赵雪春,戚飞虎,王磊。一种用于声带振动图像自动分割的自适应可变形模型。红外与毫米波学报,1999,18(5):351~357。
    [78] 夏利民,谷士文,沈新权。基于活动轮廓的多分辨率自适应图像分割。小型微型计算机系统,2001,22(2):161~164。
    [79] 贾春光,谭鸥,段会龙,吕维雪。基于变形轮廓的医学图象匹配方法。计算机辅助设计与图形学学报,1999,11(2):115~119。
    [80] 黄海赟,戚飞虎,赵雪春。基于多尺度图像的主动轮廓线模型。计算机研究与发展,2000,37(10):1240~1245。
    [81] 周彦博,张志广,杨福生。显微图像中可变形物体动态跟踪方法的研究。中国生物医学工程学报,1998,17(2):167~172。
    [82] 周激流,张晔,郭晶,朱辉。基于活动轮廓模型的人脸特征提取方法的研究。中国图象图形学报,2000,5(4):341~344。
    
    
    [83] 夏利民,谷士文。摹于形变模型的3D表面自适应重建。中国图象图形学报,2000,5(5):396~400。
    [84] 夏利民,谷士文。基于活动轮廓的运动目标的动态分割。中国图象图形学报,1999,4(8):631~634。
    [85] 黄志彦,曹春红,徐剑。基于小波多尺度分析与GVFSnake的空中目标提取。电光与控制,2002,9(4):32~35。
    [86] 黄福珍,苏剑波。基于Level Set方法的人脸轮廓提取与跟踪。计算机学报,2003,26(4):491~496。
    [87] 刘剑函,梁德群,王红光,田原。一种具有形状约束的快速Snake类算法。中国图象图形学报,1999,4(8):644~648。
    [88] M.Sonka V.Hlavac, and R. Boyle. Image Processing, Analysis, and Machine Vision(图象处理、分析与机器视觉)。北京:人民邮电出版社,2002。
    [89] 钱伟长。变分法及有限元。北京:科学出版社,1980。
    [90] 何斌,马天予,王运坚等。Visual C++数字图象处理。北京:人民邮电出版社,2001。
    [91] Bors.A.G, Pitas.I. Prediction and Tracking of Moving Objects in Irnage. IEEE Trans.on Image Processing, 2000, 9(8): 1441~1445.
    [92] Kalafatic.Z, Ribaric.S. Real-time object tracking based on optical flow and active rays. 10th Mediterranean Electrotechnical Conference, 2000,Vol Ⅱ.
    [93] 郑南宁。计算机视觉与模式识别。国防工业出版社:北京,1998。
    [94] 李智勇,沈振康,杨卫平等。动态图象分析。北京:国防工业出版社,1999年。
    [95] 艾海舟,乐秀宇。面向视觉监视实时跟踪的动态背景更新方法。计算机工程与应用,2001,19:104-106。
    [96] 雷晓峰,王耀南。远程图像监控系统的智能跟踪开发。吉首大学学报(自然科学版),2001,22(3):30~33。
    [97] 刘晓冬,苏光大等。一种可视化智能户外监控系统。中国图象图形学报,2000,5(12):1024~1029。
    
    
    [98] 王栓,艾海舟等。基于差分图象的多运动目标的检测与跟踪。中国图象图形学报,1999,4(6):470~475。
    [99] 李强,杨涛。形心和匹配跟踪算法的改进。光电工程,2000,27(2):28~30。
    [100] 付永会,张风超等。一种改进的基于颜色直方图的实时目标跟踪算法。数据采集与处理,2001,16(3):309~314。
    [101] 王春,叶虎年。虹膜识别算法的研究。贵州工业大学学报(自然科学版),2000,29(3):48~52。
    [102] 康浩,徐国治。虹膜识别系统。电路与系统学报,2000,5(1):11~15。
    [103] 严民军,齐翔林,汪云九。虹膜识别中的一种神经算法的研究。生物物理学报,2000.16(4):711~717。
    [104] Daugman.J.G. High confidence visual recognition of persons by a test of statistical independence. IEEE Trans. on Pattern and Machine Intelligence, 1993, 15:1148~1161.
    [105] 何家峰,廖曙铮,叶虎年,李柱。虹膜定位。中国图象图形学报,2000,5(3):253~255。
    [106] 李庆嵘,马争。虹膜定位算法研究。电子科技大学学报,2002,31(1):7~9。
    [107] 黄惠芳,胡广书。虹膜识别算法的研究及实现。红外与激光工程,2002,31(5):404~409。
    [108] 王蕴红,朱勇,谭铁牛。基于虹膜识别的身份鉴别。自动化学报,2002,28(1):1~10。

© 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