基于单目视觉的摄像机定位技术研究
|
摘要
摄影图像因其包含了丰富的环境信息而受到了普遍的关注,基于单目视觉的定位技术已成为了研究的热点。现有的基于单目视觉的摄像机定位方法的一般步骤是先通过图像与模板之间对应点的匹配关系建立约束得到单应矩阵,然后由单应矩阵求解摄像机的位置参数。但是图像间对应点以及图像与模板之间对应点的匹配一直是图像处理和分析的难点,匹配的正确性直接影响定位的精度。因此,在定位过程中尽量减少或者尽量避免匹配是很有实用价值的。
鉴于以上问题,本文提出了两种摄像机定位方法:基于单应矩阵的摄像机定位方法和基于图像纹理的摄像机定位方法。
基于单应矩阵的摄像机定位方法中,本文设计了一幅简单可行的模板,通过在任意方位摄取一幅模板图像建立约束,即可定位摄像机。该方法极大的减少了图像与模板之间对应点的匹配问题,可以实现连续的自主定位。
基于图像纹理的摄像机定位方法摒弃了传统的复杂矩阵变换,通过对图像纹理的分析,应用三角几何知识求解摄像机的外参数,即可对摄像机进行定位。与基于单应矩阵的摄像机定位方法相比,该方法完全摆脱了匹配问题,可以实现连续的自主定位。
实验表明,本文提出的两种方法定位精确度较高,鲁棒性强,可以满足虚拟现实,机器人导航等应用的需要。
For plenty of information of the environment, Image have been universal attended. But correspondences between points on the model plane and image one is a nodus in image process all along. Correctness of corresponding straightly influences the location precise.
In order to find more flexible camera location algorithm, two camera location technique based on computer Image Process was brought forward: one based on homography, and anther base on texture feature.
In the location technique based on homography, we designed a much simple model. The main point of the proposed technique is that it decreases the correspondences between points on the model plane and image one. And it only needs one image of the model and doesn’t move the camera. Experiments indicate that the method is reasonable, robust and accurate.
Compared to the location technique above, we can do camera location from texture feature of the picture. This proposed technique only requires the camera to observe a new designed model and it only needs one image of the model and doesn’t move the camera. The main point of this proposed technique is that it does not need to know the correspondences between points on the model plane and image one. Experiments indicate that the method is reasonable and accurate.
Experiments indicate that, the methods we put forward are reasonable, robust and accurate. And they can be used in virtual reality and Robots...
鉴于以上问题,本文提出了两种摄像机定位方法:基于单应矩阵的摄像机定位方法和基于图像纹理的摄像机定位方法。
基于单应矩阵的摄像机定位方法中,本文设计了一幅简单可行的模板,通过在任意方位摄取一幅模板图像建立约束,即可定位摄像机。该方法极大的减少了图像与模板之间对应点的匹配问题,可以实现连续的自主定位。
基于图像纹理的摄像机定位方法摒弃了传统的复杂矩阵变换,通过对图像纹理的分析,应用三角几何知识求解摄像机的外参数,即可对摄像机进行定位。与基于单应矩阵的摄像机定位方法相比,该方法完全摆脱了匹配问题,可以实现连续的自主定位。
实验表明,本文提出的两种方法定位精确度较高,鲁棒性强,可以满足虚拟现实,机器人导航等应用的需要。
For plenty of information of the environment, Image have been universal attended. But correspondences between points on the model plane and image one is a nodus in image process all along. Correctness of corresponding straightly influences the location precise.
In order to find more flexible camera location algorithm, two camera location technique based on computer Image Process was brought forward: one based on homography, and anther base on texture feature.
In the location technique based on homography, we designed a much simple model. The main point of the proposed technique is that it decreases the correspondences between points on the model plane and image one. And it only needs one image of the model and doesn’t move the camera. Experiments indicate that the method is reasonable, robust and accurate.
Compared to the location technique above, we can do camera location from texture feature of the picture. This proposed technique only requires the camera to observe a new designed model and it only needs one image of the model and doesn’t move the camera. The main point of this proposed technique is that it does not need to know the correspondences between points on the model plane and image one. Experiments indicate that the method is reasonable and accurate.
Experiments indicate that, the methods we put forward are reasonable, robust and accurate. And they can be used in virtual reality and Robots...
引文
[1] 吴卫国,陈辉堂,基于彩色图像的移动机器人定位,机器人,1999,21(5):340～346。
[2] 魏芳,董再励,用于移动机器人的视觉全局定位系统研究,机器人,2001,23(5):400～420。
[3] Shakunaga T,An Object Pose Estimation System Using a Single Camera ,IEEE Int Conf Intelligent Robots and System’s,1992:1053～1060。
[4] 马颂德,张正友,计算机视觉——计算理论与算法基础[M],北京,科学出版社,1998。
[5] R. Y. Tsai,An Efficient and Accurate Camera Calibration Technique for 3D Machine Vision,Proc. CVPR’86,New York:EEE,1986:364～374。
[6] R.O.Duda,P.E.Hart,Use of the Hough transformation to detect lines and curves in pictures,Communication of the ACM,1972,15(1):11～15。
[7] John Immerker,Some remarks on the straight line Hough transform,Patern Recognition Letters, 1998,19:1133～1135。
[8] 崔凤奎,王晓强,二值图像细化算法的比较与改进,洛阳工学院学报,1997,18(4):48～52。
[9] Lam Louisa,Lee Seong-Whan,Thinning Methodologies-A Comprehensive Survey, IEEE Trans On Pattern Analysis and Machine Intelligence,1992,14(9):869～885
[10] 何斌,马天予,Visual C++ 数字图像处理 (第二版), 北京,人民邮电出版社,2002。
[11] 李新征,易建强,赵冬斌,基于视觉的机器人定位精度提高方法,计算机测量与控制,2005.13(6):545～547。
[12] Zhengyou Zhang,A flexible new technique for camera calibration [J],IEEE Transactions on Pattern Analysis and Machine Intelligence,2000,22(11): 1330～1334。
[13] 章毓晋,图像工程(下册)——图像理解和计算机视觉,北京,清华大学出版社,1999。
[14] 于起峰,孙祥一,邱志强,从单站光测图像确定空间目标三维姿态,光学技术,2002,28(1):77～82。
[15] 胡占义,吴福朝,基于主动视觉摄像机标定方法[J],计算机学报,2002,25(11):1149～1156。
[16] 孟小桥,胡占义,摄像机自标定方法的研究与进展[J],自动化学报,2003,29(1):110～124 。
[17] 雷成,胡占义,吴福朝,一种新的基于Kruppa方程的摄像机自标定方法[J],计算机学报,2003(6):587～597。
[18] 邱茂林,马颂德,李毅,计算机视觉中摄像机定标综述[J],自动化学报,2000,26 (1):43～55,1149～1156。
[19] 李介谷,计算机视觉的理论与实践,上海,上海交通大学出版社,1991。
[20] 薛程,王士同,一种新的不基于Hough 变换的随机椭圆检测算法,微计算机信息,2006 22(1-1):265～268。
[21] 刘国翌,李华,平板图案的跟踪及其在增强现实中的应用,计算机应用,2005,25(7):1005～1008。
[22] 祝世平,强锡富,用于摄像机定位的单目视觉方法研究,光学学报,2001,21(3):450～454。
[23] 程玉仁,刘诗斌,严家明,基于数字图像处理的摄像机跟踪系统,中国有线电视,2004 (07):74～76。
[24] 于起峰,孙祥一,陈国军,用光测图像确定空间目标俯仰角和偏航角的中轴线法, 国防科技大学学报,2000,22(2):16～19。
[25] 黄桂平,李广云,单目视觉测量技术研究,计量学报,2004,25(4):315～317。
[26] 罗亚波,陈定方,采用射影几何方法实现视觉模拟系统,武汉理工大学学报,2002,24(8):83～86。
[27] 罗飞路,傅恩锡,CCD 摄像机内外部参数快速准确的标定方法,湖南大学学报,1997,24(2):71～74。
[28] Pollefeys M,Koch R,Van Gool L,Self-calibration and metric reconstruction in spite of varying and unknown internal camera parameters [A],Proceedings of the 6th International Conference on Computer Vision,1998,90～95。
[29] 杨敏,沈春林,基于单平面模板的摄像机定标研究,数据采集与处理,2003,18(1):40～43。
[30] 林倞,王涌天,刘越,基于图像匹配的户外环境注册算法,中国图像图形学报,2005,10(9):1146～1151。
[31] B .Triggs. Camera Pose and Calibration from 4 or 5 Known 3D points. In Int. Conf. Computer Vision, 1999: 278-284
[32] D. Nister. An efficient solution to the five-point relativepose problem. In IEEE Computer Vision and Pattern Recognition. 2003( II):195-202
[33] 傅德胜,寿益禾,图像图形处理学,南京,东南大学出版社,2001。
[34] 孙即祥,现代模式识别,长沙,国防科技大学出版社,2002。
[35] F. Lu. E. Milios. Robot Pose Estimation in Unknown Environments by Matching 2D Range Scans. J. Intelligent Robot. Systems, 1998. 23 (2): 267-276
[36] C. P. Lu. G. Hager, and E. Mjolsness. Fast and Globally Convergent Pose Estimation From Video Images. IEEE Trans. Pattern Analysis and Machine Intelligence. 2000, 22(5): 610-622
[37] C. S. Chen and W. Y. Chang. Pose Estmation for Generalized Imaging Device via Solving Non-Perspective N Point Problem. In. IEEE Conf. Robotics and Automation. Washing. DC, 2001:2931-2937
[2] 魏芳,董再励,用于移动机器人的视觉全局定位系统研究,机器人,2001,23(5):400～420。
[3] Shakunaga T,An Object Pose Estimation System Using a Single Camera ,IEEE Int Conf Intelligent Robots and System’s,1992:1053～1060。
[4] 马颂德,张正友,计算机视觉——计算理论与算法基础[M],北京,科学出版社,1998。
[5] R. Y. Tsai,An Efficient and Accurate Camera Calibration Technique for 3D Machine Vision,Proc. CVPR’86,New York:EEE,1986:364～374。
[6] R.O.Duda,P.E.Hart,Use of the Hough transformation to detect lines and curves in pictures,Communication of the ACM,1972,15(1):11～15。
[7] John Immerker,Some remarks on the straight line Hough transform,Patern Recognition Letters, 1998,19:1133～1135。
[8] 崔凤奎,王晓强,二值图像细化算法的比较与改进,洛阳工学院学报,1997,18(4):48～52。
[9] Lam Louisa,Lee Seong-Whan,Thinning Methodologies-A Comprehensive Survey, IEEE Trans On Pattern Analysis and Machine Intelligence,1992,14(9):869～885
[10] 何斌,马天予,Visual C++ 数字图像处理 (第二版), 北京,人民邮电出版社,2002。
[11] 李新征,易建强,赵冬斌,基于视觉的机器人定位精度提高方法,计算机测量与控制,2005.13(6):545～547。
[12] Zhengyou Zhang,A flexible new technique for camera calibration [J],IEEE Transactions on Pattern Analysis and Machine Intelligence,2000,22(11): 1330～1334。
[13] 章毓晋,图像工程(下册)——图像理解和计算机视觉,北京,清华大学出版社,1999。
[14] 于起峰,孙祥一,邱志强,从单站光测图像确定空间目标三维姿态,光学技术,2002,28(1):77～82。
[15] 胡占义,吴福朝,基于主动视觉摄像机标定方法[J],计算机学报,2002,25(11):1149～1156。
[16] 孟小桥,胡占义,摄像机自标定方法的研究与进展[J],自动化学报,2003,29(1):110～124 。
[17] 雷成,胡占义,吴福朝,一种新的基于Kruppa方程的摄像机自标定方法[J],计算机学报,2003(6):587～597。
[18] 邱茂林,马颂德,李毅,计算机视觉中摄像机定标综述[J],自动化学报,2000,26 (1):43～55,1149～1156。
[19] 李介谷,计算机视觉的理论与实践,上海,上海交通大学出版社,1991。
[20] 薛程,王士同,一种新的不基于Hough 变换的随机椭圆检测算法,微计算机信息,2006 22(1-1):265～268。
[21] 刘国翌,李华,平板图案的跟踪及其在增强现实中的应用,计算机应用,2005,25(7):1005～1008。
[22] 祝世平,强锡富,用于摄像机定位的单目视觉方法研究,光学学报,2001,21(3):450～454。
[23] 程玉仁,刘诗斌,严家明,基于数字图像处理的摄像机跟踪系统,中国有线电视,2004 (07):74～76。
[24] 于起峰,孙祥一,陈国军,用光测图像确定空间目标俯仰角和偏航角的中轴线法, 国防科技大学学报,2000,22(2):16～19。
[25] 黄桂平,李广云,单目视觉测量技术研究,计量学报,2004,25(4):315～317。
[26] 罗亚波,陈定方,采用射影几何方法实现视觉模拟系统,武汉理工大学学报,2002,24(8):83～86。
[27] 罗飞路,傅恩锡,CCD 摄像机内外部参数快速准确的标定方法,湖南大学学报,1997,24(2):71～74。
[28] Pollefeys M,Koch R,Van Gool L,Self-calibration and metric reconstruction in spite of varying and unknown internal camera parameters [A],Proceedings of the 6th International Conference on Computer Vision,1998,90～95。
[29] 杨敏,沈春林,基于单平面模板的摄像机定标研究,数据采集与处理,2003,18(1):40～43。
[30] 林倞,王涌天,刘越,基于图像匹配的户外环境注册算法,中国图像图形学报,2005,10(9):1146～1151。
[31] B .Triggs. Camera Pose and Calibration from 4 or 5 Known 3D points. In Int. Conf. Computer Vision, 1999: 278-284
[32] D. Nister. An efficient solution to the five-point relativepose problem. In IEEE Computer Vision and Pattern Recognition. 2003( II):195-202
[33] 傅德胜,寿益禾,图像图形处理学,南京,东南大学出版社,2001。
[34] 孙即祥,现代模式识别,长沙,国防科技大学出版社,2002。
[35] F. Lu. E. Milios. Robot Pose Estimation in Unknown Environments by Matching 2D Range Scans. J. Intelligent Robot. Systems, 1998. 23 (2): 267-276
[36] C. P. Lu. G. Hager, and E. Mjolsness. Fast and Globally Convergent Pose Estimation From Video Images. IEEE Trans. Pattern Analysis and Machine Intelligence. 2000, 22(5): 610-622
[37] C. S. Chen and W. Y. Chang. Pose Estmation for Generalized Imaging Device via Solving Non-Perspective N Point Problem. In. IEEE Conf. Robotics and Automation. Washing. DC, 2001:2931-2937