基于立体像对的前视二维景象图生成技术研究
|
摘要
计算机视觉的一个主要研究领域在于由二维图像识别三维目标。这个问题可以表述为:如何从多视点的二维图像去恢复三维目标信息,并进而获取目标和场景的多尺度、多视点二维图像,为后续的识别提供目标的多尺度、多视点模型。本文利用目标区域的遥感影像,获取三维信息,并建立符合飞行器观测平台视点的二维成像模型,以生成不同视点和尺度的图像序列,用于与实时图匹配,从而达到提高目标识别精度的目的。
论文提出一种基于立体像对生成多尺度多视点二维景象图的方法。首先通过目标区域的立体像对提取三维信息,建立仿真空间中目标的三维模型。第二步建立了一种物方空间坐标系到VTK仿真三维场景空间坐标系的变换,即记录飞行器观测平台飞行轨迹的地心坐标系到定位仿真三维场景的平面坐标系的变换,同时在仿真三维场景中定位飞行器观测平台姿态坐标系,用3个角分量:俯仰角、偏航角、滚动角来描述相机姿态变化,从而在仿真三维场景空间中恢复目标区域和飞行器观测平台视点之间的空间位置相对关系,同时通过成像原理和空间坐标系变换的研究证明了这种方法可以模拟飞行器观测平台视点变换的二维投影关系,并通过实验证明了生成的多尺度多视点二维投影图与飞行器观测平台实时图具有极高的相似度。
飞行器的前视景象匹配/识别技术要求生成目标区域的前视基准图。针对这种技术需求,论文提供了一种生成多视点多尺度二维投影图的方法,通过模拟飞行器平台视点变化,生成二维投影图序列,用于后续的目标识别中。
事实上,由二维基准图识别三维目标的技术已经广泛应用于城市建设规划、生态环境研究、人工智能等领域。论文的方法为提供这种技术需要的二维基准图提出了一条具有可实施性的新途径。
A main task in Computer Vision is the target recognition problem from 2D image to 3D entity. The main problem exists in such field is the way to extract multi-scale and multi-visual point 2D image and offer a multi-scale and multi-visual matching template for further target recognition by using 3D information rebuilt from multi-visual point 2D image. In this paper we proposed a method to produce multi-scale and multi-visual point image sequence by two steps:3D structure rebuilt from remote-control image and the projective geometry model we built to simulate visual point transform caused by movement of observer in air vehicles.
In this paper, we set up an algorithm to produce multi-visual point images based on satellite remote-control stereo images and did necessary research and experiments involved in Digital Photogrammetry and Computer Vision.
Firstly, we built a practical 3D reconstruction algorithm based on target’s stereo image pair, then we built a 3D model of target by using VTK(Visual toolkit).
Secondly, we proposed a coordinate system transform from object space coordinates to simulation 3D space coordinates, which means the transform from Geocentric coordinate system which determines the location of observer in air vehicles to the plane coordinate system which determines the location of target’s 3D model in the simulation 3D space. In this paper we also transformed and located the observer attitude coordinate system which described the camera pose by 3 angle parameters: elevation、roll、azimuth, from object space to simulation 3D space structured by OpenGL toolkit. Therefore, we rebuilt the relative location relationship between target and observer in air vehicle in the simulation 3D space and produce 2D reference images with the same visual point as observer in air vehicle.
This paper proposed a new method to produce multi-scale and multi-visual images. It’s necessary to match 2D reference image with 3D target generally used in applications in city plan and construction、eco-environmental construction planning、unpiloted technology and precision guidance technology.
论文提出一种基于立体像对生成多尺度多视点二维景象图的方法。首先通过目标区域的立体像对提取三维信息,建立仿真空间中目标的三维模型。第二步建立了一种物方空间坐标系到VTK仿真三维场景空间坐标系的变换,即记录飞行器观测平台飞行轨迹的地心坐标系到定位仿真三维场景的平面坐标系的变换,同时在仿真三维场景中定位飞行器观测平台姿态坐标系,用3个角分量:俯仰角、偏航角、滚动角来描述相机姿态变化,从而在仿真三维场景空间中恢复目标区域和飞行器观测平台视点之间的空间位置相对关系,同时通过成像原理和空间坐标系变换的研究证明了这种方法可以模拟飞行器观测平台视点变换的二维投影关系,并通过实验证明了生成的多尺度多视点二维投影图与飞行器观测平台实时图具有极高的相似度。
飞行器的前视景象匹配/识别技术要求生成目标区域的前视基准图。针对这种技术需求,论文提供了一种生成多视点多尺度二维投影图的方法,通过模拟飞行器平台视点变化,生成二维投影图序列,用于后续的目标识别中。
事实上,由二维基准图识别三维目标的技术已经广泛应用于城市建设规划、生态环境研究、人工智能等领域。论文的方法为提供这种技术需要的二维基准图提出了一条具有可实施性的新途径。
A main task in Computer Vision is the target recognition problem from 2D image to 3D entity. The main problem exists in such field is the way to extract multi-scale and multi-visual point 2D image and offer a multi-scale and multi-visual matching template for further target recognition by using 3D information rebuilt from multi-visual point 2D image. In this paper we proposed a method to produce multi-scale and multi-visual point image sequence by two steps:3D structure rebuilt from remote-control image and the projective geometry model we built to simulate visual point transform caused by movement of observer in air vehicles.
In this paper, we set up an algorithm to produce multi-visual point images based on satellite remote-control stereo images and did necessary research and experiments involved in Digital Photogrammetry and Computer Vision.
Firstly, we built a practical 3D reconstruction algorithm based on target’s stereo image pair, then we built a 3D model of target by using VTK(Visual toolkit).
Secondly, we proposed a coordinate system transform from object space coordinates to simulation 3D space coordinates, which means the transform from Geocentric coordinate system which determines the location of observer in air vehicles to the plane coordinate system which determines the location of target’s 3D model in the simulation 3D space. In this paper we also transformed and located the observer attitude coordinate system which described the camera pose by 3 angle parameters: elevation、roll、azimuth, from object space to simulation 3D space structured by OpenGL toolkit. Therefore, we rebuilt the relative location relationship between target and observer in air vehicle in the simulation 3D space and produce 2D reference images with the same visual point as observer in air vehicle.
This paper proposed a new method to produce multi-scale and multi-visual images. It’s necessary to match 2D reference image with 3D target generally used in applications in city plan and construction、eco-environmental construction planning、unpiloted technology and precision guidance technology.
引文
[1]张永生,巩丹超.高分辨率遥感卫星应用.第一版.北京:科学出版社, 2004. 1～11, 15～30
[2]江万寿,张剑清,张祖勋.三线阵CCD卫星影像的模拟研究.武汉大学学报, 2002, 28(4): 414～419
[3]常歌,钱曾波,黄野.城区建筑物3D景观模型建立.中国图象图形学报, 2001, 6(6): 590～593
[4]张剑清,潘励,王树根.摄影测量学.第一版.武汉:武汉大学出版社, 2003. 106～112, 117～119
[5] Calway A D, Kriiger S A. Estimating disparity and motion using multiresolution Fourier analysis. in: IEE Colloquium on Multiresolution Modelling and Analysis. Image Processing and Computer Vision, 1995,33～38
[6]黄玉琪,牟晓辉,钱曾波.航空摄影立体像对的全自动影像匹配.测绘学院学报, 2000, 17(4): 277～283
[7]王树根,李德仁,郭泽金等.正射影像上阴影和遮蔽的信息处理方法研究.测绘信息与工程, 2004, 29(4): 1～4
[8]黄玉琪,牟晓辉,钱曾波.航空摄影立体像对的全自动影像匹配.测绘学院学报, 2000, 17(4): 277～283
[9]虢建宏,田庆久.一种IKONOS影像阴影检测与去除的方法.遥感信息, 2004, 7(4): 32～35
[10] Okamoto A. Orientation and Construction of Models, Part III: Mathematical Basis of the Orientation Problem of One-dimensional Central-Perspective Photographs PE &R S, 1981, 47(12): 739-752
[11] Okamoto A, Orientation Theory of CCD Line-scanner Image International Archives of ISPRS, 1988, 27(B3): 609-617
[12] Okamoto A, ksamtsu S. Orientation Theory for Satellite CCD Line Scanner Imagery of Moan-tainous Terrain. International Archives of ISPRS, 1992, 29(B2): 205-209
[13] Okamoto A, kamatsu S. Orientation Theory for Satellite CCD Line Scanner Imageryof Hilly Terrain. International Archives of ISPRS, 19926, 29(B2): 217～221
[14] Susumu Hattori, Tetsu Ono, Clive Fraser. Orientation of High resolution Satellite Images Based on Affine Projection, International Archives of ISPRS 2000 Gingress, 2000, 33 (133): 359～366
[15] Tetsu Ono, Susumu Hattori, Hiroyuki Hasegawa, et al. Digital Mapping Using High Resolution Satellite Imagery Based on 2D Affine Projection Model. International Archives of ISPRS 2000 Congress, 2000, 33(133): 672～677
[16]张剑清,张祖勋.高分辨率遥感影像基于仿射变换的严格几何模型.武汉大学学报信息科学版, 2002, 27(6): 555～559
[17] Yasser El Manadili, Kurt Novak. Precision Rectification of SPOT Imagery Using the Direct Linear Transformation Model. PE&RS, 1996, 135～46
[18] Younian Wang, Automated triangulation of linear scanner imagery. Proceedings of ISPRS Work Groups I/1, I/3, IV/4 on“Sensors and Mapping from Space 1999”, Hanover, Germany. 1999
[19] Xinghe Yang. Accuracy of Rational Function Approximation In Photogrammetry ASPRS 2000 Annual Conference Proceedings(CD-ROM)[DB/CD]. Washington D C : American Society for Photogrammetry and Remote Sensing (ASPRS), 2000
[20] Xinghe Yang. 2001.. Piece-wise Linear rational function approximate in digital photogrammetry. ASPRS 2001 Annual Conference Proceedings. St. Louis, USA
[21] Dowman I. High resolution satellite data: Status and Issuses, Proceedings of ISPRS Working Groups I/2, I/5 and I/7 on High Resolution Mapping from Space 2001. Hanover, Germany
[22] Yong Hu, Tao C Vincent. Updating solutions of rational function model using additional control points for enhanced photogrammetric processing. In Proceedings of ISPRS Working Groups I/2, I/5 and I/7 on High Resolution Mapping from Space 2001. Hanover, Germany, 2001, 19～21, 243～251
[23]张永生贡丹超等.高分辨率遥感卫星影像的应用----成像模型、处理算法及应用技术.科学出版社, 2003. 47～49
[24]李志林朱庆.数字高程模型第二版.武汉:武汉大学出版社, 2003. 7-10
[25]朱庆陈楚江.不规则三角网的快速建立及其动态更新.武汉测绘科技大学学报, 1998, 23(3): 204-207
[26] Tsai, Victor J D. Delaunay triangulation in TIN creation: an overview and a linear-time algorithm, International Journal of GIS, 1993, 7(6): 501-524
[27] Petrie G and Kennie T. Teerain Modelling in Surveying and Civil Engineering, Computer-Aided Design, 1987, 19(4): 171-187
[28]巩丹超,张永生,邓雪清.线阵推扫影像的核线模型研究.遥感学报, 2004, 8(2): 97～101
[29] Kim T. A Study on the Epipoiarity of Linear Pushbroom Images[n]. Photogrammmetric Engnieering and Remote Sensing, 2000, (8): 44～50
[30] Brockelbank D C, Tam AP. Stereo Elevation Determination Techniques for SPOT Imagery, [J]. Photogrammetric Engineering & Remote Sensing, 1991, 57(8): 1065～1073
[31]陈鹰.遥感影像的数字摄影测量.第一版.上海:同济大学出版社, 2003. 116～127, 130～131
[32]张祖勋,周月琴.用拟合法进行SPOT影像的近似核线排列.武汉测绘科技大学学报, 1989, 14(2): 20～24
[33] F. P.普雷帕拉塔, M. I.沙莫斯著,计算几何导论,科学出版社, 1990, 56-78
[34]金通洸,王国瑾.曲面离散造型的切割磨光法.高校应用数学学报, 1988, 7(3): 5～15
[35] LEE D T. Two Algorithms for Constructing a Delaunay Triangulation[J]. Int J of Computer and Information Science, 1980, 9(3): 219-242
[36] Waston, D. F. Computing the n-Dimensional Delaunay Tessellation with Application to Voronoi Polytopes, The Computer Journal, 1981, 24(2), 167-171
[37]唐泽圣,徐志强,二维点集三角剖分的动态生成及修改.计算机辅助设计与图形学学报, 1990, 2(3): 1-8
[38] Bowyer A. Computing dirichlet tessellations. The Computer Journal, 1981, 24(2): 162～166
[39] Hartley R. In defense of the eight-point algorithm [J]. IEEE Transactions on Pattern Analysis and Intelligence, 1997, 19(6): 580-593
[40]郭修煌.精确制导技术[M].第一版.北京:国防工业出版社, 1999. 55-79
[41]孟秀云.导弹制导与控制系统原理[M].第一版.北京:北京理工大学出版社, 2003. 23-77
[42]冯宝红,王庆,万德钧. GPS车载导航中的坐标转换.中国惯性技术学报, 2002, 10(6): 29～33
[43]苏国中郑顺义张剑清张祖勋. OpenGL模拟摄影测量方法研究中国图像图形学报, 2006 11(4), 540-544
[44]巩丹超,张永生,邓雪清.线阵推扫影像的核线模型研究.遥感学报, 2004, 8(2): 97～101
[45] T. Thierry, and C. Philip. Demystification of IKONOS[J]. Earth Observation Magazine, 2000, 9(7): 17～21
[46] Taylor, C. & Kriegman, D. Structure and motion from line segments in multiple images. IEEE Trans. on Pattern Analysis and Machine Intelligence, 1995, 17(11): 1021～1032
[47] C. Baillard and A. Zisserman. Automatic reconstruction of piecewise planar models from multiple views. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, 1999, (6): 559～565
[48] Shum H, Han M, Szeliski R. Interactive construction of 3d models from panoramic mosaics [A]. CA, USA, Santa Barbara, Proceedings of the Conference on Computer Vision and Pattern Recognition [C]. 1998, 427～433
[49] T. Thierry, and C. Philip. DEM with Stereo IKONOS: A Reality Earth Observation Magazine, 2001, (7): 27～31
[50]巩丹超.高分辩率卫星遥感立体影像处理模型与算法: [博士学位论文].郑州:中国人民解放军信息工程大学, 2003
[51] Gerlach F. Characteristics of Space Imaging's One-Meter Resolution Satellite Imagery Products [A]. Int. Archives of Photogrammetry and Remote Sensing[C], Amsterdam, 2000, XXXIII(B1): 128～135
[52]文沃根.高分辨率IKONOS卫星影像及其产品的特性.遥感信息, 2001(1): 37～38
[53]史文中,曹辉,张剑清.基于高分辨率影像的城市三维建模.武汉大学学报, 2004, 29(9): 783～787
[2]江万寿,张剑清,张祖勋.三线阵CCD卫星影像的模拟研究.武汉大学学报, 2002, 28(4): 414～419
[3]常歌,钱曾波,黄野.城区建筑物3D景观模型建立.中国图象图形学报, 2001, 6(6): 590～593
[4]张剑清,潘励,王树根.摄影测量学.第一版.武汉:武汉大学出版社, 2003. 106～112, 117～119
[5] Calway A D, Kriiger S A. Estimating disparity and motion using multiresolution Fourier analysis. in: IEE Colloquium on Multiresolution Modelling and Analysis. Image Processing and Computer Vision, 1995,33～38
[6]黄玉琪,牟晓辉,钱曾波.航空摄影立体像对的全自动影像匹配.测绘学院学报, 2000, 17(4): 277～283
[7]王树根,李德仁,郭泽金等.正射影像上阴影和遮蔽的信息处理方法研究.测绘信息与工程, 2004, 29(4): 1～4
[8]黄玉琪,牟晓辉,钱曾波.航空摄影立体像对的全自动影像匹配.测绘学院学报, 2000, 17(4): 277～283
[9]虢建宏,田庆久.一种IKONOS影像阴影检测与去除的方法.遥感信息, 2004, 7(4): 32～35
[10] Okamoto A. Orientation and Construction of Models, Part III: Mathematical Basis of the Orientation Problem of One-dimensional Central-Perspective Photographs PE &R S, 1981, 47(12): 739-752
[11] Okamoto A, Orientation Theory of CCD Line-scanner Image International Archives of ISPRS, 1988, 27(B3): 609-617
[12] Okamoto A, ksamtsu S. Orientation Theory for Satellite CCD Line Scanner Imagery of Moan-tainous Terrain. International Archives of ISPRS, 1992, 29(B2): 205-209
[13] Okamoto A, kamatsu S. Orientation Theory for Satellite CCD Line Scanner Imageryof Hilly Terrain. International Archives of ISPRS, 19926, 29(B2): 217～221
[14] Susumu Hattori, Tetsu Ono, Clive Fraser. Orientation of High resolution Satellite Images Based on Affine Projection, International Archives of ISPRS 2000 Gingress, 2000, 33 (133): 359～366
[15] Tetsu Ono, Susumu Hattori, Hiroyuki Hasegawa, et al. Digital Mapping Using High Resolution Satellite Imagery Based on 2D Affine Projection Model. International Archives of ISPRS 2000 Congress, 2000, 33(133): 672～677
[16]张剑清,张祖勋.高分辨率遥感影像基于仿射变换的严格几何模型.武汉大学学报信息科学版, 2002, 27(6): 555～559
[17] Yasser El Manadili, Kurt Novak. Precision Rectification of SPOT Imagery Using the Direct Linear Transformation Model. PE&RS, 1996, 135～46
[18] Younian Wang, Automated triangulation of linear scanner imagery. Proceedings of ISPRS Work Groups I/1, I/3, IV/4 on“Sensors and Mapping from Space 1999”, Hanover, Germany. 1999
[19] Xinghe Yang. Accuracy of Rational Function Approximation In Photogrammetry ASPRS 2000 Annual Conference Proceedings(CD-ROM)[DB/CD]. Washington D C : American Society for Photogrammetry and Remote Sensing (ASPRS), 2000
[20] Xinghe Yang. 2001.. Piece-wise Linear rational function approximate in digital photogrammetry. ASPRS 2001 Annual Conference Proceedings. St. Louis, USA
[21] Dowman I. High resolution satellite data: Status and Issuses, Proceedings of ISPRS Working Groups I/2, I/5 and I/7 on High Resolution Mapping from Space 2001. Hanover, Germany
[22] Yong Hu, Tao C Vincent. Updating solutions of rational function model using additional control points for enhanced photogrammetric processing. In Proceedings of ISPRS Working Groups I/2, I/5 and I/7 on High Resolution Mapping from Space 2001. Hanover, Germany, 2001, 19～21, 243～251
[23]张永生贡丹超等.高分辨率遥感卫星影像的应用----成像模型、处理算法及应用技术.科学出版社, 2003. 47～49
[24]李志林朱庆.数字高程模型第二版.武汉:武汉大学出版社, 2003. 7-10
[25]朱庆陈楚江.不规则三角网的快速建立及其动态更新.武汉测绘科技大学学报, 1998, 23(3): 204-207
[26] Tsai, Victor J D. Delaunay triangulation in TIN creation: an overview and a linear-time algorithm, International Journal of GIS, 1993, 7(6): 501-524
[27] Petrie G and Kennie T. Teerain Modelling in Surveying and Civil Engineering, Computer-Aided Design, 1987, 19(4): 171-187
[28]巩丹超,张永生,邓雪清.线阵推扫影像的核线模型研究.遥感学报, 2004, 8(2): 97～101
[29] Kim T. A Study on the Epipoiarity of Linear Pushbroom Images[n]. Photogrammmetric Engnieering and Remote Sensing, 2000, (8): 44～50
[30] Brockelbank D C, Tam AP. Stereo Elevation Determination Techniques for SPOT Imagery, [J]. Photogrammetric Engineering & Remote Sensing, 1991, 57(8): 1065～1073
[31]陈鹰.遥感影像的数字摄影测量.第一版.上海:同济大学出版社, 2003. 116～127, 130～131
[32]张祖勋,周月琴.用拟合法进行SPOT影像的近似核线排列.武汉测绘科技大学学报, 1989, 14(2): 20～24
[33] F. P.普雷帕拉塔, M. I.沙莫斯著,计算几何导论,科学出版社, 1990, 56-78
[34]金通洸,王国瑾.曲面离散造型的切割磨光法.高校应用数学学报, 1988, 7(3): 5～15
[35] LEE D T. Two Algorithms for Constructing a Delaunay Triangulation[J]. Int J of Computer and Information Science, 1980, 9(3): 219-242
[36] Waston, D. F. Computing the n-Dimensional Delaunay Tessellation with Application to Voronoi Polytopes, The Computer Journal, 1981, 24(2), 167-171
[37]唐泽圣,徐志强,二维点集三角剖分的动态生成及修改.计算机辅助设计与图形学学报, 1990, 2(3): 1-8
[38] Bowyer A. Computing dirichlet tessellations. The Computer Journal, 1981, 24(2): 162～166
[39] Hartley R. In defense of the eight-point algorithm [J]. IEEE Transactions on Pattern Analysis and Intelligence, 1997, 19(6): 580-593
[40]郭修煌.精确制导技术[M].第一版.北京:国防工业出版社, 1999. 55-79
[41]孟秀云.导弹制导与控制系统原理[M].第一版.北京:北京理工大学出版社, 2003. 23-77
[42]冯宝红,王庆,万德钧. GPS车载导航中的坐标转换.中国惯性技术学报, 2002, 10(6): 29～33
[43]苏国中郑顺义张剑清张祖勋. OpenGL模拟摄影测量方法研究中国图像图形学报, 2006 11(4), 540-544
[44]巩丹超,张永生,邓雪清.线阵推扫影像的核线模型研究.遥感学报, 2004, 8(2): 97～101
[45] T. Thierry, and C. Philip. Demystification of IKONOS[J]. Earth Observation Magazine, 2000, 9(7): 17～21
[46] Taylor, C. & Kriegman, D. Structure and motion from line segments in multiple images. IEEE Trans. on Pattern Analysis and Machine Intelligence, 1995, 17(11): 1021～1032
[47] C. Baillard and A. Zisserman. Automatic reconstruction of piecewise planar models from multiple views. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, 1999, (6): 559～565
[48] Shum H, Han M, Szeliski R. Interactive construction of 3d models from panoramic mosaics [A]. CA, USA, Santa Barbara, Proceedings of the Conference on Computer Vision and Pattern Recognition [C]. 1998, 427～433
[49] T. Thierry, and C. Philip. DEM with Stereo IKONOS: A Reality Earth Observation Magazine, 2001, (7): 27～31
[50]巩丹超.高分辩率卫星遥感立体影像处理模型与算法: [博士学位论文].郑州:中国人民解放军信息工程大学, 2003
[51] Gerlach F. Characteristics of Space Imaging's One-Meter Resolution Satellite Imagery Products [A]. Int. Archives of Photogrammetry and Remote Sensing[C], Amsterdam, 2000, XXXIII(B1): 128～135
[52]文沃根.高分辨率IKONOS卫星影像及其产品的特性.遥感信息, 2001(1): 37～38
[53]史文中,曹辉,张剑清.基于高分辨率影像的城市三维建模.武汉大学学报, 2004, 29(9): 783～787