古建筑三维重建中的深度图像配准技术研究与应用
|
摘要
随着世界文化遗址保护工作的发展,利用计算机辅助遗址保护的相关技术成为近年来计算机可视化技术研究的热点,其中数字化是保护和推广文化遗址的主要方法之一。由于文化遗址的多样性和复杂性,遗址数字化的准确性依然难以保证。
本文主要研究利用激光扫描仪获取的深度数据对古建筑进行数字化过程中的三维数据配准技术。以小雁塔的三维虚拟数字化为研究对象,用激光扫描仪获取了小雁塔场景的点云数据,研究主要集中在点云数据预处理、多次扫描的深度图像配准、点云数据建模等方面,主要研究内容如下:
1)点云数据预处理。研究了三维点云数据的组织方式,对所获得的大场景点云数据采用了随机采样和曲率采样混合的方法进行了简化,并利用中值滤波剔除奇异点,在保证原始数据可靠性的基础上有效的降低了数据量。
2)深度图像分割。采用了Stamos提出的分割算法,用最小二乘法对点云数据进行局部平面拟合,利用共面性和同向性合并相似的平面,得到了小雁塔的深度图像分割结果,在此基础上采用了基于平面特征的方法进行了初步配准。
3)深度图像配准。研究和分析了经典ICP算法,针对古建筑的结构特征对该算法进行了改进,引入了平面特征的概念,并采用基于平方距离函数的改进ICP算法对初配准的结果迭代求精,实现了小雁塔深度图像的精确配准。
4)点云数据建模。通过对几种常见的三维重建算法的研究与比较,对简化并且配准的点云数据采用Delaunay三角剖分的方法进行了网格化,得到了比较逼真的三维模型。
As the development of protection of world cultural sites, the technology of computer-assisted sites protection has been the hotspot of the Computer Visualization Technology study. The digital is one of the major methods of preserving and promoting the culture sites. Due to the variety and complexity of the sites, it is still difficult to guarantee the accuracy in the digital of sites.
Paper is mainly on the study of 3D virtual digital SMALL WILD GOOSE PAGODA, acquiring the point cloud data from 3D laser scanner. Research focused on point cloud data preprocessing, multi-view registration, point cloud data modeling aspects. Main contents are as follows:
1) The pretreatment of point cloud data. Lots of researches of the organization of 3D point cloud data have been done. We have simplified the scenes point cloud data by the mixture method of random sampling and curvature sampling, and eliminated the singularities by using the median filtering, which reduced the volume of data effectively on the basis of assurance for the reliability of the original data.
2) The segmentation of range image. The segmentation method brought forward by Stamos has been chosen by fitting the partial plane of point by the method of the least squares fitting and combining the similitude plane based on the co-normality and co-planarity. We have achieved the segmentation results of the range images and registered them on the basis of planar feature primarily.
3) The registration of range image. We have studied and analyzed the classical ICP arithmetic, and improved it aiming at the structure features of the ancient architecture by relating the definition of planar feature and iterating the result of first-step-registration by the improved ICP arithmetic based on the square distance function, which implement the accurate registration of the range images of the SMALL WILD GOOSE PAGODA.
4) The reconstruction of point cloud data. The paper also has generated meshes by using method of Delaunay triangulation from the point cloud data simplified and registered and achieved the realistic 3D models, through the comparisons and researches of several common methods of 3D reconstruction.
本文主要研究利用激光扫描仪获取的深度数据对古建筑进行数字化过程中的三维数据配准技术。以小雁塔的三维虚拟数字化为研究对象,用激光扫描仪获取了小雁塔场景的点云数据,研究主要集中在点云数据预处理、多次扫描的深度图像配准、点云数据建模等方面,主要研究内容如下:
1)点云数据预处理。研究了三维点云数据的组织方式,对所获得的大场景点云数据采用了随机采样和曲率采样混合的方法进行了简化,并利用中值滤波剔除奇异点,在保证原始数据可靠性的基础上有效的降低了数据量。
2)深度图像分割。采用了Stamos提出的分割算法,用最小二乘法对点云数据进行局部平面拟合,利用共面性和同向性合并相似的平面,得到了小雁塔的深度图像分割结果,在此基础上采用了基于平面特征的方法进行了初步配准。
3)深度图像配准。研究和分析了经典ICP算法,针对古建筑的结构特征对该算法进行了改进,引入了平面特征的概念,并采用基于平方距离函数的改进ICP算法对初配准的结果迭代求精,实现了小雁塔深度图像的精确配准。
4)点云数据建模。通过对几种常见的三维重建算法的研究与比较,对简化并且配准的点云数据采用Delaunay三角剖分的方法进行了网格化,得到了比较逼真的三维模型。
As the development of protection of world cultural sites, the technology of computer-assisted sites protection has been the hotspot of the Computer Visualization Technology study. The digital is one of the major methods of preserving and promoting the culture sites. Due to the variety and complexity of the sites, it is still difficult to guarantee the accuracy in the digital of sites.
Paper is mainly on the study of 3D virtual digital SMALL WILD GOOSE PAGODA, acquiring the point cloud data from 3D laser scanner. Research focused on point cloud data preprocessing, multi-view registration, point cloud data modeling aspects. Main contents are as follows:
1) The pretreatment of point cloud data. Lots of researches of the organization of 3D point cloud data have been done. We have simplified the scenes point cloud data by the mixture method of random sampling and curvature sampling, and eliminated the singularities by using the median filtering, which reduced the volume of data effectively on the basis of assurance for the reliability of the original data.
2) The segmentation of range image. The segmentation method brought forward by Stamos has been chosen by fitting the partial plane of point by the method of the least squares fitting and combining the similitude plane based on the co-normality and co-planarity. We have achieved the segmentation results of the range images and registered them on the basis of planar feature primarily.
3) The registration of range image. We have studied and analyzed the classical ICP arithmetic, and improved it aiming at the structure features of the ancient architecture by relating the definition of planar feature and iterating the result of first-step-registration by the improved ICP arithmetic based on the square distance function, which implement the accurate registration of the range images of the SMALL WILD GOOSE PAGODA.
4) The reconstruction of point cloud data. The paper also has generated meshes by using method of Delaunay triangulation from the point cloud data simplified and registered and achieved the realistic 3D models, through the comparisons and researches of several common methods of 3D reconstruction.
引文
[1]Besl P J,McKay N D.A method for registration of 3-D shapes[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1992,14(2):239-256.
[2]张丽艳,逆向工程中模型重建关键技术研究[D]。南京:南京航空航天大学,2001。
[3]柯映林,离散数据的几何造型技术及其应用研究[D]。南京:南京航空航天大学,1992。
[4]G.Barequet and M.Sharir.Partial Surface matching by using directed footprints.Computational Geometry:Theory and Applications,Volume 12,Numberl-2,February 1999,45-62.
[5]K.S Arun,T.S Huang,S.D Blostein.Least square fitting of two 3D point sets.IEEE Transactions on Pattern Analysis and Machine Intelligence,1987,9(5):698-700.
[6]Chen Y,Medioni G.Object modeling by registration of multiple range images[J].Image and Vision Computing,1992,10(3):145-155.
[7]Potmesil M.Generating models of solid objects by matching 3D surface segments[c].Proceedings of the 8th International Joint conference on Artificial Intelligence,Karlsruhe,1983:1089-1093.
[8]Masuda T,Yokoya N.A robust method for registration and segmentation of multiple range images[J].Computer Vision and Image Understanding 1995,61(3):295-307.
[9]Johnson A,Hebert M.Surface registration by matching oriented points[C].Proceedings of International Conference on Recent Advances in 3-D Digital Imaging and Modeling,Ottawa,1997:121-128.
[10]罗先波,钟约先,李仁举,三维扫描系统中的数据配准技术[J]。清华大学学报:自然科学版,2004,44(8):1104-1106。
[11]王磊,刑渊,反向工程中数据点云的拼合[J]。模具技术,2004(1)47-49。
[12]Emanuele Trucco,Andrea Fusiello.Robust motion and correspondence of noisy 3-D point sets with missing data.Pattern Recognition Letters 20(9):889-898,1999.
[13]Marcos A.Rodrigues and Yonghuai Liu.Exploiting Structural Constraints for Accurate Image Regitration.Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems(IROS 2002),September 30-October 4,2002,Lausanne,Switzerland,vol.1,pp:1-6.
[14]何文峰,大型场景三维重建中的深度图像配准[D]。北京大学,硕士论文,2004:7-10。
[15]Fruh,Christian.Automated 3D model generation for urban environments:[dissertation].Karlsruhe,Gemany:Univ.of Karlsruhe,2002.
[16]张爱武,基于激光扫描数据的室外场景表面重建方法[J]。首都师范大学,系统仿真学报,Vol.17 No.2,Feb.2005。
[17]戴静兰,海量点云预处理算法研究[D]。浙江大学.,硕士论文,2006:41-44。
[18]O.D.Faugeras and M.Hebert.The representation,recognition,and locating of 3d objects.International Journal of Robotic Research,vol.5,no.3,pp.27-52,Fall 1986.
[19]F.Stein and G.Medioni.Structural indexing:efficient 3-D object recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence,vol.14,no.2,pp.125.
[20]I.Stamos and P.K.Allen.Geometry and Texture Recovery of Scenes of Large Scale.Computer Vision and Image Understanding(CVIU),vol.88,no.2,Nov.2002,pp.94-118.
[21]Paul J.Besl and Ramesh C.Jain.Segmentation through variable-order surface fitting.IEEE Transactions on Pattern Analysis and Machine Intelligence,Vol.10,No.2,pp.167-192,March 1988.
[22]R.L.Hoffman and A.K.Jain.Segmentation and classification of range images.IEEE Transactions on Pattern Analysis and Machine Intelligence,vol.9,no.5,pp.608-620,1987.
[23]I.Stamos,P.E.Allen,3-D model construction using range and image data.In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2000(CVPR'00),vol,1,pp.531-536.
[24]G.Sharp,S.Lee,D.Wehe.ICP registration using Invariant Features.Transactions on Pattern Analysis and Machine Intelligece.2002,24(1):90-102.
[25]Lin.Z,T.S.Huang,S.D.Blostein,H.Lee,E.A.Margerum.Motion Estimation from 3D Point Sets with and without Correspondence.Proceedings of International Conference on Computer Vision and Pattern Recognition,Miami,1986:194-201.
[26]J.K.Aggarwal,M.J.Magee.Determining Motion Parameters using Intensity Guided Range Sensing.Pattern Recognition.1986(19):169-180.
[27]O.D.Faugeras,Q.-T.Luong,S.J.Maybank.Camera Self-Calibration:Theory and Experiments.European Conference on Computer Vision,Santa Matgherita Ligure,1992: 321-334.
[28]B.K.P.Horn.Closed-form Solution of Absolute Orientation using unit quaternion.Journal of the Optical Society of America.1987,4(4):629-642.
[29]戴静兰,陈志杨,叶修梓。ICP算法在点云配准中的应用[J],中国图象图形学报,2007
[30]J.Luck,C.Little,W.Hoff.Registration of Range Data using a Hybrid Simulated Annealing and Iterative Closest Point Algorithm.Robotics and Automation.Proceedings ICRA'00.2000,vol.4,3739-3744.
[31]K.Pulli.Multiview Registration for Large Data Sets.Proceedings of International Conference on 3D Digital Imaging and Modeling,Ottawa,1999,National Research Council of Canada,1999:160-168.
[32]M.A.Rodriguez,Liu Y.H.,Wei.Q.Geometric Alignment of Two Overlapping Range Images.International Conference on Acoustics,Speech,and Signal Processing.2001,vol.3,1689-1692.
[33]K.Masuda,K,Sakaue,N.Yokoya.Registration and Integration of Multiple Range Images for 3D Model Construction.Proceedings of the 13th International Conference on Pattern Recognition.Vienna,1996:879-883.
[34]K.Nishino,K.Ikeuchi.Robust Simultaneous Registration of Multiple Range Images.Proceedings of Asian Conference on Computer Vision,Melbourne,2002:454-461.
[35]H.Gagnon,M.Soucy,R.Bergevin.Registration of Multiple Range Views for Automatic 3D Model Building.Proceedings of International Conference on Computer Vision and Pattern Recognition,Seattle,1994:581-586.
[36]吴剑光,基于深度图像的三维模型重建研究[D]。哈尔滨工业大学,硕士论文,2004。
[37]Blais G,Levine M D.Registering multiview range data to create 3d computer graphics.IEEE Transaction on Pattern Analysis and Machine Intelligence,1995,17(8):820-824.
[38]Li Q,Griffiths J G.Interative closest geometric objects registration.Computer and Mathematics with Applications,2000,40:1171-1118.
[39]张学昌,习俊通,严隽琪.基于点云数据的复杂型面数字化检测技术研究[J]。计算机集成制造系统,2005。
[40]胡寅,三维扫描仪与逆向工程关键技术研究[D],华中科技大学,博士论文,2005:79-83。
[41]H.Hoppe,T.DeRose,T.Duchamp,J.McDonald and W.Stuetzle,Surface reconstruction from unorganized points,Computer Graphics,1992,26(2):71-76.
[42]N.Amenta,M.Bern,M.Kamvysselis,A new Voronoi-based surface reconstruction algorithm,SIGGRAPH'98,415-421.
[43]C.Bradley,Rapid protyoping models generated form machine vision data Computers in Industry,2001,41,159-173.
[44]M.S.Floater and M.Riemers,Meshless parameterization and surface reconstruction,Computer Aided Geometric Design,2001,18,77-92.
[45]成欣,周明全,耿国华.空间三角网格曲面的补洞方法[J],计算机应用研究,2006。
[46]伍毅,三维扫描信息获取的深度图像配准算法设计及开发[D]。浙江大学,硕士论文,2005,7-15。
[2]张丽艳,逆向工程中模型重建关键技术研究[D]。南京:南京航空航天大学,2001。
[3]柯映林,离散数据的几何造型技术及其应用研究[D]。南京:南京航空航天大学,1992。
[4]G.Barequet and M.Sharir.Partial Surface matching by using directed footprints.Computational Geometry:Theory and Applications,Volume 12,Numberl-2,February 1999,45-62.
[5]K.S Arun,T.S Huang,S.D Blostein.Least square fitting of two 3D point sets.IEEE Transactions on Pattern Analysis and Machine Intelligence,1987,9(5):698-700.
[6]Chen Y,Medioni G.Object modeling by registration of multiple range images[J].Image and Vision Computing,1992,10(3):145-155.
[7]Potmesil M.Generating models of solid objects by matching 3D surface segments[c].Proceedings of the 8th International Joint conference on Artificial Intelligence,Karlsruhe,1983:1089-1093.
[8]Masuda T,Yokoya N.A robust method for registration and segmentation of multiple range images[J].Computer Vision and Image Understanding 1995,61(3):295-307.
[9]Johnson A,Hebert M.Surface registration by matching oriented points[C].Proceedings of International Conference on Recent Advances in 3-D Digital Imaging and Modeling,Ottawa,1997:121-128.
[10]罗先波,钟约先,李仁举,三维扫描系统中的数据配准技术[J]。清华大学学报:自然科学版,2004,44(8):1104-1106。
[11]王磊,刑渊,反向工程中数据点云的拼合[J]。模具技术,2004(1)47-49。
[12]Emanuele Trucco,Andrea Fusiello.Robust motion and correspondence of noisy 3-D point sets with missing data.Pattern Recognition Letters 20(9):889-898,1999.
[13]Marcos A.Rodrigues and Yonghuai Liu.Exploiting Structural Constraints for Accurate Image Regitration.Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems(IROS 2002),September 30-October 4,2002,Lausanne,Switzerland,vol.1,pp:1-6.
[14]何文峰,大型场景三维重建中的深度图像配准[D]。北京大学,硕士论文,2004:7-10。
[15]Fruh,Christian.Automated 3D model generation for urban environments:[dissertation].Karlsruhe,Gemany:Univ.of Karlsruhe,2002.
[16]张爱武,基于激光扫描数据的室外场景表面重建方法[J]。首都师范大学,系统仿真学报,Vol.17 No.2,Feb.2005。
[17]戴静兰,海量点云预处理算法研究[D]。浙江大学.,硕士论文,2006:41-44。
[18]O.D.Faugeras and M.Hebert.The representation,recognition,and locating of 3d objects.International Journal of Robotic Research,vol.5,no.3,pp.27-52,Fall 1986.
[19]F.Stein and G.Medioni.Structural indexing:efficient 3-D object recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence,vol.14,no.2,pp.125.
[20]I.Stamos and P.K.Allen.Geometry and Texture Recovery of Scenes of Large Scale.Computer Vision and Image Understanding(CVIU),vol.88,no.2,Nov.2002,pp.94-118.
[21]Paul J.Besl and Ramesh C.Jain.Segmentation through variable-order surface fitting.IEEE Transactions on Pattern Analysis and Machine Intelligence,Vol.10,No.2,pp.167-192,March 1988.
[22]R.L.Hoffman and A.K.Jain.Segmentation and classification of range images.IEEE Transactions on Pattern Analysis and Machine Intelligence,vol.9,no.5,pp.608-620,1987.
[23]I.Stamos,P.E.Allen,3-D model construction using range and image data.In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2000(CVPR'00),vol,1,pp.531-536.
[24]G.Sharp,S.Lee,D.Wehe.ICP registration using Invariant Features.Transactions on Pattern Analysis and Machine Intelligece.2002,24(1):90-102.
[25]Lin.Z,T.S.Huang,S.D.Blostein,H.Lee,E.A.Margerum.Motion Estimation from 3D Point Sets with and without Correspondence.Proceedings of International Conference on Computer Vision and Pattern Recognition,Miami,1986:194-201.
[26]J.K.Aggarwal,M.J.Magee.Determining Motion Parameters using Intensity Guided Range Sensing.Pattern Recognition.1986(19):169-180.
[27]O.D.Faugeras,Q.-T.Luong,S.J.Maybank.Camera Self-Calibration:Theory and Experiments.European Conference on Computer Vision,Santa Matgherita Ligure,1992: 321-334.
[28]B.K.P.Horn.Closed-form Solution of Absolute Orientation using unit quaternion.Journal of the Optical Society of America.1987,4(4):629-642.
[29]戴静兰,陈志杨,叶修梓。ICP算法在点云配准中的应用[J],中国图象图形学报,2007
[30]J.Luck,C.Little,W.Hoff.Registration of Range Data using a Hybrid Simulated Annealing and Iterative Closest Point Algorithm.Robotics and Automation.Proceedings ICRA'00.2000,vol.4,3739-3744.
[31]K.Pulli.Multiview Registration for Large Data Sets.Proceedings of International Conference on 3D Digital Imaging and Modeling,Ottawa,1999,National Research Council of Canada,1999:160-168.
[32]M.A.Rodriguez,Liu Y.H.,Wei.Q.Geometric Alignment of Two Overlapping Range Images.International Conference on Acoustics,Speech,and Signal Processing.2001,vol.3,1689-1692.
[33]K.Masuda,K,Sakaue,N.Yokoya.Registration and Integration of Multiple Range Images for 3D Model Construction.Proceedings of the 13th International Conference on Pattern Recognition.Vienna,1996:879-883.
[34]K.Nishino,K.Ikeuchi.Robust Simultaneous Registration of Multiple Range Images.Proceedings of Asian Conference on Computer Vision,Melbourne,2002:454-461.
[35]H.Gagnon,M.Soucy,R.Bergevin.Registration of Multiple Range Views for Automatic 3D Model Building.Proceedings of International Conference on Computer Vision and Pattern Recognition,Seattle,1994:581-586.
[36]吴剑光,基于深度图像的三维模型重建研究[D]。哈尔滨工业大学,硕士论文,2004。
[37]Blais G,Levine M D.Registering multiview range data to create 3d computer graphics.IEEE Transaction on Pattern Analysis and Machine Intelligence,1995,17(8):820-824.
[38]Li Q,Griffiths J G.Interative closest geometric objects registration.Computer and Mathematics with Applications,2000,40:1171-1118.
[39]张学昌,习俊通,严隽琪.基于点云数据的复杂型面数字化检测技术研究[J]。计算机集成制造系统,2005。
[40]胡寅,三维扫描仪与逆向工程关键技术研究[D],华中科技大学,博士论文,2005:79-83。
[41]H.Hoppe,T.DeRose,T.Duchamp,J.McDonald and W.Stuetzle,Surface reconstruction from unorganized points,Computer Graphics,1992,26(2):71-76.
[42]N.Amenta,M.Bern,M.Kamvysselis,A new Voronoi-based surface reconstruction algorithm,SIGGRAPH'98,415-421.
[43]C.Bradley,Rapid protyoping models generated form machine vision data Computers in Industry,2001,41,159-173.
[44]M.S.Floater and M.Riemers,Meshless parameterization and surface reconstruction,Computer Aided Geometric Design,2001,18,77-92.
[45]成欣,周明全,耿国华.空间三角网格曲面的补洞方法[J],计算机应用研究,2006。
[46]伍毅,三维扫描信息获取的深度图像配准算法设计及开发[D]。浙江大学,硕士论文,2005,7-15。