机器人视觉目标跟踪方法的研究
|
摘要
在智能机器人系统中视觉系统是一个重要的技术组成部分，它为智能机器人提供最丰富的位置参数和运动参数等信息，以方便智能机器人能够正确感知自身所处的工作环境，进一步对自身的行为做出正确的决策。视觉跟踪技术是视觉研究的一个重要分支，在智能机器人的科学研究、工程应用以及竞技娱乐等方面都发挥着重要的作用。它是一种通过图像处理的方法获取目标参数并对目标进行跟踪的视觉技术。
在对特定目标的视觉跟踪过程中提出一种基于运动目标颜色特征和形状特征的目标跟踪方法。该方法首先在图像全局范围内搜索与目标颜色特征匹配的区域，然后对这些区域采用不变矩特征匹配运动目标的形状，如果颜色和形状都匹配的目标总数大于总目标数，则采用预测模型筛选出运动目标所在的正确区域，并最终确定目标的位置。试验验证了该方法对具有固定形状和颜色的特定目标有良好的跟踪能力和全局跟踪特性。
针对彩色图像HSI颜色模型各分量所携带的颜色信息不同的特点，结合区域分割算法在彩色图像中提取区域颜色特征作为目标的颜色特征，它包含一维颜色特征值和区域灰度均值。一维颜色特征值具有较好的稳定性，不随光照强度改变而变化，反映目标内在的颜色特征；区域灰度均值是连通区域各像素灰度的平均值，它反映同一连通区域的亮度。试验数据显示，当区域灰度均值在一定范围内变化时，一维颜色特征值仍然具有较好的稳定性。
区域分割技术是区域颜色特征提取过程中必须采用的技术，提出一种结合梯度算子和子区合并技术的区域分割方法。该方法采用梯度算子对整幅图像进行梯度运算，并将运算结果二值化，然后采用子区合并技术连通具有相似性的相邻区域完成区域分割。通过梯度运算和二值化过程，在一定程度上消除了同一区域中颜色值变化对分割结果的影响，分割试验结果表明只要相邻像素的颜色值变化范围小于二值化阈值，这种分割方法就能够实现对图像的有效分割。
The vision system is one of the important parts in the intelligent robot system, it provides much information about the position and the kinematics for intelligent robots. The intelligent robot may apperceive the working environment and make decision for behavior properly according to the information provided by the vision system. The visual tracking technology is one of branches of visual technologies. It plays important actions in some aspects such as research, application and entertainment of intelligent robots. It extracts parameters of object by employing methods of image processing, and then tracks the object.
An object tracking method based on the color features and the shape features of object is presented for tracking a particular object. The method searches some regions whose color features match the object's in the whole image, and then it matches the shape features of object among the regions ascertained by the first step. If the number of all regions whose color and shape features all match the object's is bigger than the total number of objects which will be tracked, the real regions in which the object locates are filtered by using prediction model, so the positions of the object is made sure. The experiments validate the ability of the method on tracking the color and rigid-shape object, this method has global tracking trait.
Region color features which are extracted from color image according to the different color information of the different components in HSI color model, act as the color features of the object. They include One Dimension Color Eigenvalue and Region Mean Gray Scale. One Dimension Color Eigenvalue has good stability, it is not changed when the luminance is changed, so it reflects the natural color features of objects. Region Mean Gray Scale is the mean gray scale value of all pixels in one connective region, it reflects the luminance of the region. The experiment data show that One Dimension Color Eigenvalue is steady when Region Mean Gray Scale is changed at some extent.
Region segmentation technology must be used during the process of
在对特定目标的视觉跟踪过程中提出一种基于运动目标颜色特征和形状特征的目标跟踪方法。该方法首先在图像全局范围内搜索与目标颜色特征匹配的区域，然后对这些区域采用不变矩特征匹配运动目标的形状，如果颜色和形状都匹配的目标总数大于总目标数，则采用预测模型筛选出运动目标所在的正确区域，并最终确定目标的位置。试验验证了该方法对具有固定形状和颜色的特定目标有良好的跟踪能力和全局跟踪特性。
针对彩色图像HSI颜色模型各分量所携带的颜色信息不同的特点，结合区域分割算法在彩色图像中提取区域颜色特征作为目标的颜色特征，它包含一维颜色特征值和区域灰度均值。一维颜色特征值具有较好的稳定性，不随光照强度改变而变化，反映目标内在的颜色特征；区域灰度均值是连通区域各像素灰度的平均值，它反映同一连通区域的亮度。试验数据显示，当区域灰度均值在一定范围内变化时，一维颜色特征值仍然具有较好的稳定性。
区域分割技术是区域颜色特征提取过程中必须采用的技术，提出一种结合梯度算子和子区合并技术的区域分割方法。该方法采用梯度算子对整幅图像进行梯度运算，并将运算结果二值化，然后采用子区合并技术连通具有相似性的相邻区域完成区域分割。通过梯度运算和二值化过程，在一定程度上消除了同一区域中颜色值变化对分割结果的影响，分割试验结果表明只要相邻像素的颜色值变化范围小于二值化阈值，这种分割方法就能够实现对图像的有效分割。
The vision system is one of the important parts in the intelligent robot system, it provides much information about the position and the kinematics for intelligent robots. The intelligent robot may apperceive the working environment and make decision for behavior properly according to the information provided by the vision system. The visual tracking technology is one of branches of visual technologies. It plays important actions in some aspects such as research, application and entertainment of intelligent robots. It extracts parameters of object by employing methods of image processing, and then tracks the object.
An object tracking method based on the color features and the shape features of object is presented for tracking a particular object. The method searches some regions whose color features match the object's in the whole image, and then it matches the shape features of object among the regions ascertained by the first step. If the number of all regions whose color and shape features all match the object's is bigger than the total number of objects which will be tracked, the real regions in which the object locates are filtered by using prediction model, so the positions of the object is made sure. The experiments validate the ability of the method on tracking the color and rigid-shape object, this method has global tracking trait.
Region color features which are extracted from color image according to the different color information of the different components in HSI color model, act as the color features of the object. They include One Dimension Color Eigenvalue and Region Mean Gray Scale. One Dimension Color Eigenvalue has good stability, it is not changed when the luminance is changed, so it reflects the natural color features of objects. Region Mean Gray Scale is the mean gray scale value of all pixels in one connective region, it reflects the luminance of the region. The experiment data show that One Dimension Color Eigenvalue is steady when Region Mean Gray Scale is changed at some extent.
Region segmentation technology must be used during the process of
引文
[1] 张广军编著．机器视觉．第一版．北京：科学出版社，2005：1-2页，3页，5-6页
[2] 谢存禧，张铁等编著．机器人技术及其应用．第一版．北京：机械工业出版社，2005：120-121页，123页
[3] 陈利红．机器人视觉系统的研究与开发．浙江大学硕士学位论文．2003：3页
[4] 刘兴荣．用于移动机器人的立体视觉技术．河北工业大学硕士学位论文．2003：2页
[5] Tissainayagam P, Suter D. Contour trackingwith automatic motion model switching. Pattern Recognition. 2003, 36(10): 2411-2427P
[6] Ponsa D, Roca X. Multiple model to deformable shape tracking. Francisco J P, Aurelio J C C, Nicolas P et al. Pattern Recognition and Image Analysis: First Iberian Conference, IbPRIA 2003, Puerto de Andratx, Mallorca, Spain, 2003: 782-792P
[7] Carmichael O, Hebert M. Object recognition by a cascade of edge probes. British Machine Vision Conference 2002, British Machine Vision Association, 2002, 1: 103-112P
[8] Zhong Y, Jain A, Dubuisson J M. Object tracking using deformable templates. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2000, 22(5): 544-549P
[9] Peterfreund N. Robust tracking of position and velocity with Kalman snakes. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1999, 21(6): 564-569P
[10] Tissainayagam P, Suter D. Object tracking in image sequences using point features. Pattern Recognition. 2005, 38: 105-113P
[11] 邱兆文，张田文．一种新的图像颜色特征提取方法．哈尔滨工业大学学报．2004，36(12)：1699-1701页
[12] 冀亚丽，程小平．基于重量化多特征提取的彩色图像检索．西南师范大学??学报．2005，30(2)：254-258页
[13] Microsoft Corporation. Microsoft DirectX 9.0 SDK
[14] 陆其明．DirectShow系列讲座之一——DirectShow系统概述．http://hqtech.nease.net/Article/DS Lecturel.htm．2005．6．27
[15] 陆其明．FiIter Graph的事件交互机制．http://hqtech.nease.net/Article/DS_EventNotify.htm．2005．6．27
[16] 陆其明．DirectShow系列讲座之二——Filter原理．http://hqtech.nease.net/Article/DS_Lecture2.htm．2005．6．27
[17] 陆其明．深入浅出DirectShow FiIter．http://hqtech.nease.net/Article/DS_ThinkingInDS.htm．2005．6．27
[18] 陆其明．DirectShow对硬件的支持原理．http://hqtech.nease.net/Article/DS_HardSupport.htm．2005．6．27
[19] 沈浩，谭骏珊．DirectShow流媒体信息获取及图象转换．http://www.yesky.com/259/1854259.shtml．2005．6．27
[20] Chen T Q，Lu Y．Color image segmentation——an innovative approach. Pattern Recognition.2002, 35: 395-405P
[21] Wang H Z, Suter D. Color image segmentation using global information and local homogeneity. In Proceedings 7th International Conference on Digital Image Computing，Techniques and Applications, Sydney, 2003: 89-98P
[22] 陈科庆，何茂军．彩色图像分割综述．湖北师范学院学报．2004，24(4)：32-36页
[23] 徐志节，杨杰，王猛．一种新的彩色图像降维方法．上海交通大学学报．2004，38(12)：2063-2067页
[24] Mehtre B M, Kankanhalli M S, Lee W F. Content-based image retrieval using a composite color-shape approach. Information Processing and Management. 1998, 34(1): 109-120P
[25] Qiu G P. Embedded colour image coding for content-based retrieval. Journal of Visual Communication and Image Representation. 2004, 15(4): 507-521P[26] Nikolaou N, Papamarkos N. Color image retrieval using a fractal signature extraction technique. Engineering Applications of Artificial Intelligence. 2002, 15(1): 80-96P
[27] 欧宗瑛，袁野，张艳珍．基于颜色信息足球机器人视觉跟踪算法．大连理工大学学报．2000，40(6)：729-732页
[28] 于秀芬．轮式移动机器人视觉定位与目标跟踪研究．南京航空航天大学硕士学位论文．2004
[29] 周俊．农用轮式移动机器人视觉导航系统的研究．南京农业大学博士学位论文．2003
[30] 李庆忠，陈显华，顾伟康．基于彩色立体视觉的障碍物快速检测方法．计算机科学．2003，30(9)：72-75页
[31] Chiang C C, Huang C J. A robust method for detecting arbitrarily tilted human faces in color images. Pattern Recognition Letters. 2005, 26(16): 2518-2536P
[32] Kapfer M, Benois-Pineau J. Detection of human faces in color image sequences with arbitrary motions for very low bit-rate videophone coding. Pattern Recognition Letters. 1997, 18(14): 1503-1518P
[33] Wang J G, Sung E. Frontal-view face detection and facial feature extraction using color and morphological operations. Pattern Recognition Letters. 1999, 20(10): 1053-1068P
[34] 王耀明编著．图像的矩函数——原理、算法及应用．第一版．上海：华东理工大学出版社，2002：3页
[35] Liu J, Zhang T X. Fast algorithm for generation of moment invariants. Pattern Recognition. 2004, 37(8): 1745-1756P
[36] Wong W H, Siu W C, Lam K M. Generation of moment invariants and their uses for character recognition. Pattern Recognition Letters. 1995, 16(2): 115-123P
[37] Shao Yuan, Celenk M. Higher-order spectra (HOS) invariants for shape recognition. Pattern Recognition. 2001, 34(11): 2097-2113P
[38] Chong C Way, Raveendran P, Mukundan R. Translation invariants of??Zernike moments. Pattern Recognition. 2003, 36(8): 1765-1773P
[39] Chong C Way, Raveendran P, Mukundan R. Translation and scale invariants of Legendre moments. Pattern Recognition. 2004, 37(1): 119-129P
[40] Gonzalez R C, Woods R E. Digital Image Processing. 2nd ed. Beijing: Publishing House of Electronics Industry, 2002: 299P
[41] 皮文凯．基于全方位视觉的人体运动检测和跟踪．北京大学硕士学位论文．2004
[42] 董宁．基于移动机器人全方位视觉的人体运动检测．北京大学硕士学位论文．2005
[43] Noyer J C, Lanvin P, Benjelloun M. Non-iinear matched fiitering for object detection and tracking. Pattern Recognition Letters. 2004, 25(6): 655-668P
[44] 汤思维，陈卫东，曹其新．移动机器人多目标彩色视觉跟踪系统．机器人．2003，25(1)：10-14页
[45] 原新，朱齐丹，许辉．移动机器人的视觉处理方法及实现．哈尔滨工程大学学报．2003，24(1)：41-44页
[46] 于秀芬，段海滨，龚华军．移动机器人视觉定位方法的研究与实现．数据采集与处理．2004，19(4)：433-437页
[47] 彭刚，黄心汉，王敏等．基于视觉引导和超声测距的运动目标跟踪和抓取．高技术通讯．2002，12(6)：1-6页
[48] Sogo T, Ishiguro H, Trivedi M M. Real-Time target localization and tracking by N-ocular stereo. IEEE Workshop on Omnidirectional Vision (OMNIVIS' 00). 2000: 153-160P
[2] 谢存禧，张铁等编著．机器人技术及其应用．第一版．北京：机械工业出版社，2005：120-121页，123页
[3] 陈利红．机器人视觉系统的研究与开发．浙江大学硕士学位论文．2003：3页
[4] 刘兴荣．用于移动机器人的立体视觉技术．河北工业大学硕士学位论文．2003：2页
[5] Tissainayagam P, Suter D. Contour trackingwith automatic motion model switching. Pattern Recognition. 2003, 36(10): 2411-2427P
[6] Ponsa D, Roca X. Multiple model to deformable shape tracking. Francisco J P, Aurelio J C C, Nicolas P et al. Pattern Recognition and Image Analysis: First Iberian Conference, IbPRIA 2003, Puerto de Andratx, Mallorca, Spain, 2003: 782-792P
[7] Carmichael O, Hebert M. Object recognition by a cascade of edge probes. British Machine Vision Conference 2002, British Machine Vision Association, 2002, 1: 103-112P
[8] Zhong Y, Jain A, Dubuisson J M. Object tracking using deformable templates. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2000, 22(5): 544-549P
[9] Peterfreund N. Robust tracking of position and velocity with Kalman snakes. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1999, 21(6): 564-569P
[10] Tissainayagam P, Suter D. Object tracking in image sequences using point features. Pattern Recognition. 2005, 38: 105-113P
[11] 邱兆文，张田文．一种新的图像颜色特征提取方法．哈尔滨工业大学学报．2004，36(12)：1699-1701页
[12] 冀亚丽，程小平．基于重量化多特征提取的彩色图像检索．西南师范大学??学报．2005，30(2)：254-258页
[13] Microsoft Corporation. Microsoft DirectX 9.0 SDK
[14] 陆其明．DirectShow系列讲座之一——DirectShow系统概述．http://hqtech.nease.net/Article/DS Lecturel.htm．2005．6．27
[15] 陆其明．FiIter Graph的事件交互机制．http://hqtech.nease.net/Article/DS_EventNotify.htm．2005．6．27
[16] 陆其明．DirectShow系列讲座之二——Filter原理．http://hqtech.nease.net/Article/DS_Lecture2.htm．2005．6．27
[17] 陆其明．深入浅出DirectShow FiIter．http://hqtech.nease.net/Article/DS_ThinkingInDS.htm．2005．6．27
[18] 陆其明．DirectShow对硬件的支持原理．http://hqtech.nease.net/Article/DS_HardSupport.htm．2005．6．27
[19] 沈浩，谭骏珊．DirectShow流媒体信息获取及图象转换．http://www.yesky.com/259/1854259.shtml．2005．6．27
[20] Chen T Q，Lu Y．Color image segmentation——an innovative approach. Pattern Recognition.2002, 35: 395-405P
[21] Wang H Z, Suter D. Color image segmentation using global information and local homogeneity. In Proceedings 7th International Conference on Digital Image Computing，Techniques and Applications, Sydney, 2003: 89-98P
[22] 陈科庆，何茂军．彩色图像分割综述．湖北师范学院学报．2004，24(4)：32-36页
[23] 徐志节，杨杰，王猛．一种新的彩色图像降维方法．上海交通大学学报．2004，38(12)：2063-2067页
[24] Mehtre B M, Kankanhalli M S, Lee W F. Content-based image retrieval using a composite color-shape approach. Information Processing and Management. 1998, 34(1): 109-120P
[25] Qiu G P. Embedded colour image coding for content-based retrieval. Journal of Visual Communication and Image Representation. 2004, 15(4): 507-521P[26] Nikolaou N, Papamarkos N. Color image retrieval using a fractal signature extraction technique. Engineering Applications of Artificial Intelligence. 2002, 15(1): 80-96P
[27] 欧宗瑛，袁野，张艳珍．基于颜色信息足球机器人视觉跟踪算法．大连理工大学学报．2000，40(6)：729-732页
[28] 于秀芬．轮式移动机器人视觉定位与目标跟踪研究．南京航空航天大学硕士学位论文．2004
[29] 周俊．农用轮式移动机器人视觉导航系统的研究．南京农业大学博士学位论文．2003
[30] 李庆忠，陈显华，顾伟康．基于彩色立体视觉的障碍物快速检测方法．计算机科学．2003，30(9)：72-75页
[31] Chiang C C, Huang C J. A robust method for detecting arbitrarily tilted human faces in color images. Pattern Recognition Letters. 2005, 26(16): 2518-2536P
[32] Kapfer M, Benois-Pineau J. Detection of human faces in color image sequences with arbitrary motions for very low bit-rate videophone coding. Pattern Recognition Letters. 1997, 18(14): 1503-1518P
[33] Wang J G, Sung E. Frontal-view face detection and facial feature extraction using color and morphological operations. Pattern Recognition Letters. 1999, 20(10): 1053-1068P
[34] 王耀明编著．图像的矩函数——原理、算法及应用．第一版．上海：华东理工大学出版社，2002：3页
[35] Liu J, Zhang T X. Fast algorithm for generation of moment invariants. Pattern Recognition. 2004, 37(8): 1745-1756P
[36] Wong W H, Siu W C, Lam K M. Generation of moment invariants and their uses for character recognition. Pattern Recognition Letters. 1995, 16(2): 115-123P
[37] Shao Yuan, Celenk M. Higher-order spectra (HOS) invariants for shape recognition. Pattern Recognition. 2001, 34(11): 2097-2113P
[38] Chong C Way, Raveendran P, Mukundan R. Translation invariants of??Zernike moments. Pattern Recognition. 2003, 36(8): 1765-1773P
[39] Chong C Way, Raveendran P, Mukundan R. Translation and scale invariants of Legendre moments. Pattern Recognition. 2004, 37(1): 119-129P
[40] Gonzalez R C, Woods R E. Digital Image Processing. 2nd ed. Beijing: Publishing House of Electronics Industry, 2002: 299P
[41] 皮文凯．基于全方位视觉的人体运动检测和跟踪．北京大学硕士学位论文．2004
[42] 董宁．基于移动机器人全方位视觉的人体运动检测．北京大学硕士学位论文．2005
[43] Noyer J C, Lanvin P, Benjelloun M. Non-iinear matched fiitering for object detection and tracking. Pattern Recognition Letters. 2004, 25(6): 655-668P
[44] 汤思维，陈卫东，曹其新．移动机器人多目标彩色视觉跟踪系统．机器人．2003，25(1)：10-14页
[45] 原新，朱齐丹，许辉．移动机器人的视觉处理方法及实现．哈尔滨工程大学学报．2003，24(1)：41-44页
[46] 于秀芬，段海滨，龚华军．移动机器人视觉定位方法的研究与实现．数据采集与处理．2004，19(4)：433-437页
[47] 彭刚，黄心汉，王敏等．基于视觉引导和超声测距的运动目标跟踪和抓取．高技术通讯．2002，12(6)：1-6页
[48] Sogo T, Ishiguro H, Trivedi M M. Real-Time target localization and tracking by N-ocular stereo. IEEE Workshop on Omnidirectional Vision (OMNIVIS' 00). 2000: 153-160P