基于深度图像的三维重建技术研究
|
摘要
随着计算机视觉技术和计算机图形学技术的发展,三维重建技术的应用领域从机器人导航和视觉检测等高端领域扩展到了虚拟现实、视觉模拟等领域,而且在文物保护、生物医学、建筑和机械的设计与制造方面也得到了广泛的应用。基于深度图像的三维重建技术相较于传统的建模技术效率更高,省时省力,而且可以恢复被测物体的三维几何信息,重建模型具有更高的真实性。
本论文首先介绍了深度图像的概念等相关知识,以及由深度图像进行三维重建的流程,然后针对三维重建过程中的数据预处理、图像配准及三维重建技术进行了研究。本文研究内容如下：
(1)数据预处理。本文从点云去噪和数据精简两个方面对数据预处理操作进行了研究。针对无序点云数据,论文提出了基于中值的k-d tree去噪算法,该方法在去除噪声点的同时,可有效保留原始数据的细节。在精简数据部分,本文采用结合随机采样和曲率采样的混合采样方法对点云数据进行简化,既保持了原始数据的特征,也有效降低了点云数据量。
(2)深度图像的配准。深度图像的配准技术是三维重建工作中的重点。本文对遗传算法及ICP算法进行了分析研究,提出了一种新的配准思路。首先采用遗传算法进行初步配准,然后利用基于平方距离函数的ICP算法对初配准结果迭代求精。遗传算法具有强大的全局最优搜索能力,并具有较好的问题域的独立性和应用的鲁棒性,初配准可有效缩小两幅深度图像之间的位置差,提高ICP算法的稳定性。
(3)三维重建。本文分析研究了常见三维重建方法,并详细阐述了Delaunay三角剖分原理及步骤。并介绍了拉普拉斯曲面平滑算法,该算法可有效降低重建表面粗糙度,提高重建效果。
With the development of computer vision and computer graphics technologies, the application fields expanded from robot navigation, vision inspection to virtual reality, and they are be use widely in preservation of cultural relics, biomedicine, building construction, design and manufacture of machine. Compared to the traditional reconstruction, image-based reconstruction technology is faster and easier, and the 3D information of the object can be obtained, and the reconstruction models have better authenticity.
This paper introduced some knowledge about the range image and the process of the 3D reconstruction based on the range image. The pretreatment of the data, registration of the range image and the 3D reconstruction technology are emphasizes in 3D reconstruction. The contents of the paper as follows:
(1)The pretreatment of the point cloud data. This paper does some research on noise points and simplifying. The k-d tree based on median is used in random point cloud for wiping noise, and efficiency in speed and result, and the characteristics of the data are contained. The mixture method of random sampling and curvature sampling is used for simplifying. It can reduce the number of the point cloud and contain the details of the original data also.
(2) The registration of the range image.The registration is the emphasis in 3D reconstruction. The paper proposed a new "coarse-to-fine" thought of registration based on the genetic algorithm and the ICP algorithm. The paper used the genetic algorithm in rough registration, and iterating the result of the rough registration by the ICP based on the square distance function. The genetic algorithm's powerful global optimal search capabilities, the independence of the problem domain and the robust nature of its application can improve the stability of ICP algorithm.
(3)3D reconstruction. The paper analyzed several common methods of 3D reconstruction, and presented the theory and the process of the Delaunay triangulation. In the same time, the Laplacian smoothing is also introduced. It can lower the roughness, and improve the effect of the reconstruction.
本论文首先介绍了深度图像的概念等相关知识,以及由深度图像进行三维重建的流程,然后针对三维重建过程中的数据预处理、图像配准及三维重建技术进行了研究。本文研究内容如下：
(1)数据预处理。本文从点云去噪和数据精简两个方面对数据预处理操作进行了研究。针对无序点云数据,论文提出了基于中值的k-d tree去噪算法,该方法在去除噪声点的同时,可有效保留原始数据的细节。在精简数据部分,本文采用结合随机采样和曲率采样的混合采样方法对点云数据进行简化,既保持了原始数据的特征,也有效降低了点云数据量。
(2)深度图像的配准。深度图像的配准技术是三维重建工作中的重点。本文对遗传算法及ICP算法进行了分析研究,提出了一种新的配准思路。首先采用遗传算法进行初步配准,然后利用基于平方距离函数的ICP算法对初配准结果迭代求精。遗传算法具有强大的全局最优搜索能力,并具有较好的问题域的独立性和应用的鲁棒性,初配准可有效缩小两幅深度图像之间的位置差,提高ICP算法的稳定性。
(3)三维重建。本文分析研究了常见三维重建方法,并详细阐述了Delaunay三角剖分原理及步骤。并介绍了拉普拉斯曲面平滑算法,该算法可有效降低重建表面粗糙度,提高重建效果。
With the development of computer vision and computer graphics technologies, the application fields expanded from robot navigation, vision inspection to virtual reality, and they are be use widely in preservation of cultural relics, biomedicine, building construction, design and manufacture of machine. Compared to the traditional reconstruction, image-based reconstruction technology is faster and easier, and the 3D information of the object can be obtained, and the reconstruction models have better authenticity.
This paper introduced some knowledge about the range image and the process of the 3D reconstruction based on the range image. The pretreatment of the data, registration of the range image and the 3D reconstruction technology are emphasizes in 3D reconstruction. The contents of the paper as follows:
(1)The pretreatment of the point cloud data. This paper does some research on noise points and simplifying. The k-d tree based on median is used in random point cloud for wiping noise, and efficiency in speed and result, and the characteristics of the data are contained. The mixture method of random sampling and curvature sampling is used for simplifying. It can reduce the number of the point cloud and contain the details of the original data also.
(2) The registration of the range image.The registration is the emphasis in 3D reconstruction. The paper proposed a new "coarse-to-fine" thought of registration based on the genetic algorithm and the ICP algorithm. The paper used the genetic algorithm in rough registration, and iterating the result of the rough registration by the ICP based on the square distance function. The genetic algorithm's powerful global optimal search capabilities, the independence of the problem domain and the robust nature of its application can improve the stability of ICP algorithm.
(3)3D reconstruction. The paper analyzed several common methods of 3D reconstruction, and presented the theory and the process of the Delaunay triangulation. In the same time, the Laplacian smoothing is also introduced. It can lower the roughness, and improve the effect of the reconstruction.
引文
[1]王宇宙.计算机视觉三维重建理论与应用[D].西安：西北大学计算机软件与理论,2004.
[2]谢凯,汤晓安等.一种基于多幅深度图像的混合建模方法[J].小型微型计算机系统.2003,24(1)：94-96.
[3]闫雯.基于图像的三维建模算法与实验研究[D].西安：西安理工大学,2007.
[4]杨长江,汪威,胡占义.一种基于主动视觉的相机内参数白标定方法[J].计算机学报.1998.
[5]杨长江,苏凤梅,胡占义.基于平面二次曲线的摄像机标定[J].计算机学报.2000,23(5)：541-547.
[6]陈明舟,王仲.基于视觉技术的三维曲面非接触无损检测方法[J].无损检测.2001,23(9)：372-374,413.
[7]梁栋,刘春梅.基于遗传算法的射影重构[J].中国图象图形学报.2002,7(5)：445-449.
[8]刘钢,彭群生,鲍虎军.基于图像建模技术研究综述与展望[J].计算机辅助设计与图形学学报.2005,17(1)：18-27.
[9]Minoru Asada, Takamaro,T. Visual Tracking Of Unknown Moving Object By on Multisensor Fusion and Intelligent Systems.1999.
[10]舒志龙,阮秋琦.快速对称多窗口正交三目立体匹配[J].计算机研究与发展.2001,38(10)：1229-1235.
[11]Okutomi M,Kanade T.A mutiple-baseline stereo[J].IEEE Transanctions on Pattern Analysis and Machine Intelligenc.1993,15(4):353.
[12]王晓云.基于图像的建模技术[J].涪陵师范学院学报.2005,21(5)：56-58.
[13]全红艳.基于几何与图像结合的虚拟场景绘制技术研究[D].哈尔滨：哈尔滨工业大学计算机科学与技术学院,2006.
[14]季铮.近景目标三维重建的自动化方法研究[D].武汉：武汉大学,2007.
[15]A. Bartoli.The Geometry of Dynamic Scenes On Coplanar and Convergent Linear Motions Embedded in 3D Static Scenes[J].Computer Vision and Image Understanding.2005,98(2):223-238.
[16]王莉莉,刘嵘.基于图像的几何三维重建方法[J].系统仿真学报.2001,13：77-81,93.
[17]V. Kolmogorov,R.Zabih. Multi-camera Scene Reconstruction via Graph Cuts[J]. European Conefrence of Computer Vision.2002,3:82-96.
[18]陶曼.深度图像的分割与压缩[D].北京：首都师范大学,2006.
[19]周颖.深度图像的获取及其处理[D].西安：西安电子科技大学,2008.
[20]栾悉道,应龙,谢毓湘等.三维建模技术研究进展[J].计算机科学.2008,35(2)：208-210,229.
[21]张凯.三维激光扫描数据的空间配准研究[D].南京：南京师范大学,2008.
[22]董明晓,郑康平.一种点云数据噪声点的随机滤波处理方法.中国图象图形学报[J].2004,9(2)：245-248.
[23]戴静兰.海量点云预处理算法研究[D].浙江：浙江大学,2006.
[24]吕渭萍.多视点深度图像的配准方法研究与实现[D].四川：电子科技大学,2008.
[25]Weir D J, Milroy M, Bradley C, Vickers G W. Reverse engineering physical models employing wrap-around B-spline surfaces and quadrics.Proc Introduction Mech Engrs-Part B.1996,210:147-157.
[26]Martin R R, Stroud I A, Marshal A D. Data reduction for reverse engineering. RECCAD, Deliverable Document COPERUNICUS project. Computer and Automation Institute of Hungarian Academy of Science.1996(1068):63-69.
[27]Alexa M,Behr J,Cohen-Or D,Fleishman S,Levin D,Sliva T. Point Set Surfaces. IEEE Visualization.2001,01.
[28]Turk G. Re-tiling polygonal surfaces.SIGGRPAH 1992,92.
[29]陈三清.点云数据的配准方法研究[J].计算机技术与自动化.2009：52-53.
[30]马婷.点云数据的显示与配准[D].吉林：吉林大学,2007.
[31]高鹏东.深度像配准及曲面细分技术的研究与应用[D].天津：天津大学,2006.
[32]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.
[33]J. K. Aggarwal, M. J.Magee. Determining Motion Parameters using Intensity Guided Range Sensing. Patterm Recognition.1986(19):169-180.
[34]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.
[35]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.
[36]李敏强,林丹,李书全.遗传算法的基本理论与应用[D].北京：科学出版社,2002.
[37]刘正东.计算机视觉中立体匹配技术的研究[D].南京：南京理工大学,2005.
[38]尹丹.古建筑三维重建中的深度图像配准技术研究与应用[D].西安：西北大学,2008.
[39]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.
[40]Li Q, Griffiths J G.Interative closest geometric objects registration. Computer and Mathematics with Applications.2000,40:1171-1178.
[41]黄淼.散乱点云的网格重构算法研究[D].河南：河南科技大学,2008.
[42]万琳.基于三角网格的图像表示方法研究[D].华中科技大学,2009.
[43]丁帆.点云数据三维网格构造方法研究[D].武汉：华中科技大学,2007.
[2]谢凯,汤晓安等.一种基于多幅深度图像的混合建模方法[J].小型微型计算机系统.2003,24(1)：94-96.
[3]闫雯.基于图像的三维建模算法与实验研究[D].西安：西安理工大学,2007.
[4]杨长江,汪威,胡占义.一种基于主动视觉的相机内参数白标定方法[J].计算机学报.1998.
[5]杨长江,苏凤梅,胡占义.基于平面二次曲线的摄像机标定[J].计算机学报.2000,23(5)：541-547.
[6]陈明舟,王仲.基于视觉技术的三维曲面非接触无损检测方法[J].无损检测.2001,23(9)：372-374,413.
[7]梁栋,刘春梅.基于遗传算法的射影重构[J].中国图象图形学报.2002,7(5)：445-449.
[8]刘钢,彭群生,鲍虎军.基于图像建模技术研究综述与展望[J].计算机辅助设计与图形学学报.2005,17(1)：18-27.
[9]Minoru Asada, Takamaro,T. Visual Tracking Of Unknown Moving Object By on Multisensor Fusion and Intelligent Systems.1999.
[10]舒志龙,阮秋琦.快速对称多窗口正交三目立体匹配[J].计算机研究与发展.2001,38(10)：1229-1235.
[11]Okutomi M,Kanade T.A mutiple-baseline stereo[J].IEEE Transanctions on Pattern Analysis and Machine Intelligenc.1993,15(4):353.
[12]王晓云.基于图像的建模技术[J].涪陵师范学院学报.2005,21(5)：56-58.
[13]全红艳.基于几何与图像结合的虚拟场景绘制技术研究[D].哈尔滨：哈尔滨工业大学计算机科学与技术学院,2006.
[14]季铮.近景目标三维重建的自动化方法研究[D].武汉：武汉大学,2007.
[15]A. Bartoli.The Geometry of Dynamic Scenes On Coplanar and Convergent Linear Motions Embedded in 3D Static Scenes[J].Computer Vision and Image Understanding.2005,98(2):223-238.
[16]王莉莉,刘嵘.基于图像的几何三维重建方法[J].系统仿真学报.2001,13：77-81,93.
[17]V. Kolmogorov,R.Zabih. Multi-camera Scene Reconstruction via Graph Cuts[J]. European Conefrence of Computer Vision.2002,3:82-96.
[18]陶曼.深度图像的分割与压缩[D].北京：首都师范大学,2006.
[19]周颖.深度图像的获取及其处理[D].西安：西安电子科技大学,2008.
[20]栾悉道,应龙,谢毓湘等.三维建模技术研究进展[J].计算机科学.2008,35(2)：208-210,229.
[21]张凯.三维激光扫描数据的空间配准研究[D].南京：南京师范大学,2008.
[22]董明晓,郑康平.一种点云数据噪声点的随机滤波处理方法.中国图象图形学报[J].2004,9(2)：245-248.
[23]戴静兰.海量点云预处理算法研究[D].浙江：浙江大学,2006.
[24]吕渭萍.多视点深度图像的配准方法研究与实现[D].四川：电子科技大学,2008.
[25]Weir D J, Milroy M, Bradley C, Vickers G W. Reverse engineering physical models employing wrap-around B-spline surfaces and quadrics.Proc Introduction Mech Engrs-Part B.1996,210:147-157.
[26]Martin R R, Stroud I A, Marshal A D. Data reduction for reverse engineering. RECCAD, Deliverable Document COPERUNICUS project. Computer and Automation Institute of Hungarian Academy of Science.1996(1068):63-69.
[27]Alexa M,Behr J,Cohen-Or D,Fleishman S,Levin D,Sliva T. Point Set Surfaces. IEEE Visualization.2001,01.
[28]Turk G. Re-tiling polygonal surfaces.SIGGRPAH 1992,92.
[29]陈三清.点云数据的配准方法研究[J].计算机技术与自动化.2009：52-53.
[30]马婷.点云数据的显示与配准[D].吉林：吉林大学,2007.
[31]高鹏东.深度像配准及曲面细分技术的研究与应用[D].天津：天津大学,2006.
[32]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.
[33]J. K. Aggarwal, M. J.Magee. Determining Motion Parameters using Intensity Guided Range Sensing. Patterm Recognition.1986(19):169-180.
[34]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.
[35]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.
[36]李敏强,林丹,李书全.遗传算法的基本理论与应用[D].北京：科学出版社,2002.
[37]刘正东.计算机视觉中立体匹配技术的研究[D].南京：南京理工大学,2005.
[38]尹丹.古建筑三维重建中的深度图像配准技术研究与应用[D].西安：西北大学,2008.
[39]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.
[40]Li Q, Griffiths J G.Interative closest geometric objects registration. Computer and Mathematics with Applications.2000,40:1171-1178.
[41]黄淼.散乱点云的网格重构算法研究[D].河南：河南科技大学,2008.
[42]万琳.基于三角网格的图像表示方法研究[D].华中科技大学,2009.
[43]丁帆.点云数据三维网格构造方法研究[D].武汉：华中科技大学,2007.