无人机倾斜影像自动检索及影像姿态恢复
|
摘要
无人机倾斜摄影测量颠覆了以往正射影像只能从垂直角度拍摄的局限,在三维建模中有广泛的前景。针对有些无人机倾斜影像数据无相机标定参数、无航带信息(无序)、无POS信息的现状,本文以计算机视觉中基于内容的影像检索方法与改进的渐进式SFM方法为基础,提出一种"三无"影像自动检索、空中三角测量及影像三维重建的方法。该方法首先通过提取的特征检索出相似影像并建立网络结构,然后将影像进行两两匹配增强对应关系并进行连接点的追踪,最后利用光束法平差方法对其进行平差,获取影像集的三维点云,提高大规模影像检索、影像匹配速度的同时,提高重建的精确性和鲁棒度。本文选取三组典型试验区大数据量倾斜影像数据进行试验,立体实测控制点中误差可以达到平面0.16m/高程0.18m,试验验证了方法的稳定性、可靠性和实用性。
In recent years, unmanned aerial vehicle(UAV) have become a means of civilization and universalization.The UAV image is gradually replacing aerospace remote sensing data and is widely used in many fields. The limitation that the orthophotos can only be taken from a vertical angle in the past has been broken nowadays by oblique photogrammetry which has wide application prospect in 3D modeling. Aiming to ensure three-no-image(i.e., no camera calibration parameters, no strip information(disordered), and no POS(Position and orientation System) information) in some oblique images, the paper proposes a method of automatic aerial triangulation and 3D reconstruction for the three-no-image. This method is based on the content-based image retrieval method and improved progressive SFM(Structure from Motion) method in computer vision. Firstly, the method retrieves similar images and establishes the network through extracted features. Secondly, the correspondence between the two images is enhanced by matching the images and the tie points are tracked. Thirdly, the 3D point cloud of image is obtained by bundle adjustment. The algorithm improves the accuracy and robustness of reconstruction and makes a great progress in large scale image retrieval and image matching. Finally,, the stability, reliability, and accuracy of the proposed method was tested and validated with three-test experiments by using large scale real oblique images over three test areas.The test-1 area has 1190 images, from the project construction to the final aerial triangulation calculation without control,the total time is 4.3 hours, and the error is 0.4 pixels. The test-2 has 3685 images and no POS is used in experiment. From the project construction to the final aerial triangulation calculation without control, it takes 8 hours and the error is within 0.32 pixels. The two experimental results verified the stability and applicability of the proposed algorithm. The test-3 area has 1346 images, after 5 hours processing, the error of the free network adjustment is within0.42 pixels, 9 ground control points are used for check points, the error is within 0.16 m in plane and 0.18 m in elevation. The experimental results verified the accuracy and reliability of the proposed algorithm.
In recent years, unmanned aerial vehicle(UAV) have become a means of civilization and universalization.The UAV image is gradually replacing aerospace remote sensing data and is widely used in many fields. The limitation that the orthophotos can only be taken from a vertical angle in the past has been broken nowadays by oblique photogrammetry which has wide application prospect in 3D modeling. Aiming to ensure three-no-image(i.e., no camera calibration parameters, no strip information(disordered), and no POS(Position and orientation System) information) in some oblique images, the paper proposes a method of automatic aerial triangulation and 3D reconstruction for the three-no-image. This method is based on the content-based image retrieval method and improved progressive SFM(Structure from Motion) method in computer vision. Firstly, the method retrieves similar images and establishes the network through extracted features. Secondly, the correspondence between the two images is enhanced by matching the images and the tie points are tracked. Thirdly, the 3D point cloud of image is obtained by bundle adjustment. The algorithm improves the accuracy and robustness of reconstruction and makes a great progress in large scale image retrieval and image matching. Finally,, the stability, reliability, and accuracy of the proposed method was tested and validated with three-test experiments by using large scale real oblique images over three test areas.The test-1 area has 1190 images, from the project construction to the final aerial triangulation calculation without control,the total time is 4.3 hours, and the error is 0.4 pixels. The test-2 has 3685 images and no POS is used in experiment. From the project construction to the final aerial triangulation calculation without control, it takes 8 hours and the error is within 0.32 pixels. The two experimental results verified the stability and applicability of the proposed algorithm. The test-3 area has 1346 images, after 5 hours processing, the error of the free network adjustment is within0.42 pixels, 9 ground control points are used for check points, the error is within 0.16 m in plane and 0.18 m in elevation. The experimental results verified the accuracy and reliability of the proposed algorithm.
引文
[1]廖小罕,周成虎,苏奋振,等.无人机遥感众创时代[J].地球信息科学学报,2016,18(11):1439-1447.[Liao X H,Zhou C H,Su F Z,et al.The mass innovation era of UAV remote sensing[J].Journal of Geo-information Science,2016,18(11):1439-1447.]
[2]孙钰珊,张力,艾海滨,等.倾斜影像匹配与三维建模关键技术发展综述[J].遥感信息,2018,33(2):1-8.[Sun Y S,Zhang L,Ai H B,et al.Overview of key technologies in image matching and 3D reconstruction for oblique photogrammetry[J].Remote Sensing Information,2018,33(2):1-8.]
[3]陈天博,胡卓玮,魏铼,等.无人机遥感数据处理与滑坡信息提取[J].地球信息科学学报,2017,19(5):692-701.[Chen TB,Hu Z W,Wei L,et al.Data processing and landslide information extraction based on UAV remote sensing[J].Journal of Geo-information Science,2017,19(5):692-701.]
[4]王豪.三维重建中的影像检索方法研究[M].北京:中国测绘科学研究院,2018.[Wang H.Research on Image Retrieval Method in 3D Reconstruction[M].Beijing:Chinese Academy of Surveying and Mapping,2018.]
[5]查冰.基于非量测相机无结构影像的运动估计算法研究[M].北京.中国测绘科学研究院,2017.[Cha B.Algorithm research on motion estimation of unstructured image based on consumer camera[M].Beijing:Chinese Academy of Surveying and Mapping,2017.]
[6]Snavely N,Seitz S M,Szeliski R.Photo tourism:Exploring photo collections in 3D[J].International Conference on Computer Graphics and Interactive Techniques,2006,25(3):835-46.
[7]Agarwal S,Furukawa Y,Snavely N,et al.Building Rome in a day[C].Proceedings of the IEEE International Conference on Computer Vision,F,2009.
[8]Pollefeys M,R,D,Frahm J M,et al.Detailed real-time urban 3D reconstruction from video[J].International Journal of Computer Vision,2008,78(2):143-67.
[9]Wu C.Towards linear-time incremental structure from motion[C].Proceedings of the International Conference on 3dv-Conference,F,2013.
[10]Schnberger J L,Frahm J M.Structure-from-Motion Revisited[C].Proceedings of the Computer Vision and Pattern Recognition,F,2016.
[11]Moulon P,Monasse P,Marlet R.Adaptive structure from motion with a contrario model estimation[C].Proceedings of the Asian Conference on Computer Vision,F,2012.
[12]Flickner M,Sawhney H,Niblack W,et al.Query by image and video content:The QBIC system[J].IEEE Computer,2002,28(9):255-264
[13]王海龙,王佩雪.基于内容的图像检索探讨[J].中原工学院学报,2005,16(2):12-15.[Wang H L,Wang P X.Analysis of content-based image retrieval[J].Zhongyuan Institute of Technology,2005,16(2):12-15.]
[14]王欣.基于WWW的图像检索技术[J].现代图书情报技术,2002,18(3):70-73.[Wang X.WWW-Based image retrieval techniques[J].New Technology of Library and Information Service,2002,18(3):70-73.]
[15]Smith J R,Chang S.F.Visually searching the web for content[J].IEEE Multimedia,2002,4(3):12-20.
[16]祝晓斌,刘亚奇,蔡强,等.基于内容的图像检索技术研究[J].计算机仿真,2015,32(5):1-4.[Zhu X B,Liu Y Q,Cai Q.A survey on content-based image retrieval technology[J].Computer Simulation,2015,32(5):1-4.]
[17]Zhang N,Man K L,Yu T,et al.Text and content based image retrieval via locality sensitive hashing[J].Engineering Letters,2011,19(3):228-234.
[18]Rehman M,Iqbal M,Sharif M,et al.Content based image retrieval:Survey[J].Telkomnika Indonesian Journal of Electrical Engineering,2013,11(11):995-998.
[19]姜亚莉.基于内容的图像检索系统分析[J].测绘与空间地理信息,2012,35(1):119-120.[Jiang Y L.Analysis of the content-based image retrieval system[J].Geomatics&Spatial Information Technology,2012,35(1):119-120.]
[20]黄祥林,沈兰荪,蔡强,等.基于内容的图像检索技术研究[J].电子学报,2002,30(7):1065-1071.[Huang X L,Shen L S,Cai Q.Research on content-based image retrieval techniques[J].Acta Electronica Sinica,2002,30(7):1065-1071.]
[21]Lowe D G.Distinctive image features from scale-invariant Keypoints[J].International Journal of Computer Vision,2004,60(2):91-110.
[22]Kneip L,Scaramuzza D,Siegwart R.A novel parametrization of the perspective-three-point problem for a direct computation of absolute camera position and orientation[C].Proceedings of the the IEEE Conference on Computer Vision and Pattern Recognition,F,2011.
[2]孙钰珊,张力,艾海滨,等.倾斜影像匹配与三维建模关键技术发展综述[J].遥感信息,2018,33(2):1-8.[Sun Y S,Zhang L,Ai H B,et al.Overview of key technologies in image matching and 3D reconstruction for oblique photogrammetry[J].Remote Sensing Information,2018,33(2):1-8.]
[3]陈天博,胡卓玮,魏铼,等.无人机遥感数据处理与滑坡信息提取[J].地球信息科学学报,2017,19(5):692-701.[Chen TB,Hu Z W,Wei L,et al.Data processing and landslide information extraction based on UAV remote sensing[J].Journal of Geo-information Science,2017,19(5):692-701.]
[4]王豪.三维重建中的影像检索方法研究[M].北京:中国测绘科学研究院,2018.[Wang H.Research on Image Retrieval Method in 3D Reconstruction[M].Beijing:Chinese Academy of Surveying and Mapping,2018.]
[5]查冰.基于非量测相机无结构影像的运动估计算法研究[M].北京.中国测绘科学研究院,2017.[Cha B.Algorithm research on motion estimation of unstructured image based on consumer camera[M].Beijing:Chinese Academy of Surveying and Mapping,2017.]
[6]Snavely N,Seitz S M,Szeliski R.Photo tourism:Exploring photo collections in 3D[J].International Conference on Computer Graphics and Interactive Techniques,2006,25(3):835-46.
[7]Agarwal S,Furukawa Y,Snavely N,et al.Building Rome in a day[C].Proceedings of the IEEE International Conference on Computer Vision,F,2009.
[8]Pollefeys M,R,D,Frahm J M,et al.Detailed real-time urban 3D reconstruction from video[J].International Journal of Computer Vision,2008,78(2):143-67.
[9]Wu C.Towards linear-time incremental structure from motion[C].Proceedings of the International Conference on 3dv-Conference,F,2013.
[10]Schnberger J L,Frahm J M.Structure-from-Motion Revisited[C].Proceedings of the Computer Vision and Pattern Recognition,F,2016.
[11]Moulon P,Monasse P,Marlet R.Adaptive structure from motion with a contrario model estimation[C].Proceedings of the Asian Conference on Computer Vision,F,2012.
[12]Flickner M,Sawhney H,Niblack W,et al.Query by image and video content:The QBIC system[J].IEEE Computer,2002,28(9):255-264
[13]王海龙,王佩雪.基于内容的图像检索探讨[J].中原工学院学报,2005,16(2):12-15.[Wang H L,Wang P X.Analysis of content-based image retrieval[J].Zhongyuan Institute of Technology,2005,16(2):12-15.]
[14]王欣.基于WWW的图像检索技术[J].现代图书情报技术,2002,18(3):70-73.[Wang X.WWW-Based image retrieval techniques[J].New Technology of Library and Information Service,2002,18(3):70-73.]
[15]Smith J R,Chang S.F.Visually searching the web for content[J].IEEE Multimedia,2002,4(3):12-20.
[16]祝晓斌,刘亚奇,蔡强,等.基于内容的图像检索技术研究[J].计算机仿真,2015,32(5):1-4.[Zhu X B,Liu Y Q,Cai Q.A survey on content-based image retrieval technology[J].Computer Simulation,2015,32(5):1-4.]
[17]Zhang N,Man K L,Yu T,et al.Text and content based image retrieval via locality sensitive hashing[J].Engineering Letters,2011,19(3):228-234.
[18]Rehman M,Iqbal M,Sharif M,et al.Content based image retrieval:Survey[J].Telkomnika Indonesian Journal of Electrical Engineering,2013,11(11):995-998.
[19]姜亚莉.基于内容的图像检索系统分析[J].测绘与空间地理信息,2012,35(1):119-120.[Jiang Y L.Analysis of the content-based image retrieval system[J].Geomatics&Spatial Information Technology,2012,35(1):119-120.]
[20]黄祥林,沈兰荪,蔡强,等.基于内容的图像检索技术研究[J].电子学报,2002,30(7):1065-1071.[Huang X L,Shen L S,Cai Q.Research on content-based image retrieval techniques[J].Acta Electronica Sinica,2002,30(7):1065-1071.]
[21]Lowe D G.Distinctive image features from scale-invariant Keypoints[J].International Journal of Computer Vision,2004,60(2):91-110.
[22]Kneip L,Scaramuzza D,Siegwart R.A novel parametrization of the perspective-three-point problem for a direct computation of absolute camera position and orientation[C].Proceedings of the the IEEE Conference on Computer Vision and Pattern Recognition,F,2011.