多视几何无人机影像堆体体积量算
|
摘要
针对形态不规则、大规模或不便于近距离实测的堆体体积的计算问题,借助低空无人机(Unmanned Aerial Vehicle,UAV)搭载非量测的普通数码相机对堆体进行倾斜摄影,获取堆体多视角的倾斜影像。利用运动恢复结构和基于面片的多视角立体视觉(Sf M-PMVS)技术处理由无人机获取的倾斜影像。在引入地面像控点后,首先对影像进行特征点提取和基于最近邻距离比率(Nearest Neighbor Distance Ratio,NNDR)算法的SIFT粗匹配,采用随机抽样一致算法(Random Sample Consensus,RANSAC)剔除误匹配点对进而精确求得影像的基本矩阵F完成影像匹配。引入经相机检校得到的相机内参数精确求解本质矩阵E,恢复相机运动姿态后由投影矩阵P计算稀疏点云在物方坐标系下的坐标,采用PMVS算法进行点云密集匹配,经光束法平差后得到堆体在物方坐标系下精确的三维密集点云。对三维密集点云做点云分割,剔除非堆体表面点后构建Delaunay三角网,利用数字地面模型(Digital Terrain Model,DTM)法计算堆体的体积。与用GNSS-RTK均匀测得堆体表面三维坐标点采用DTM法计算堆体体积的结果对比证明,所给方法计算堆体的体积在准确性上能满足实际生产中的要求。
For the calculation of stack volume with irregular shape,large scale or inconvenient closerange measurement,this paper studies the tilt photography of the stack by using a non-measured ordinary digital camera with UAV. A tilted image of the multi-view of the stack is obtained. The tilted image acquired by the drone is processed using the structure from motion and patch-based multi-view stereo( SfM-PMVS) technique. After the introduction of the ground image control point,the feature point extraction and the SIFT rough matching based on the nearest neighbor distance ratio( NNDR) algorithm are firstly applied,and the random sample consensus( RANSAC) is used to eliminate the mismatched point and then accurately obtain the basic matrix F of the image to complete image matching. The incamera parameters introduced into the calibration are used to accurately solve the essential matrix E.After the camera pose is restored,the coordinates of the sparse point cloud in the object coordinate system are calculated by the projection matrix P,and the point cloud is closely matched by the PMVS al-gorithm. Obtain a precise three-dimensional dense point cloud of the stack in the object coordinate system,conduct point cloud segmentation on the 3 D dense point cloud,and build a Delaunay triangulation after the non-stacked surface point of the stack is eliminated. The volume of the pile is calculated by the Digital Terrain Model( DTM) method. The calculation results of this method are compared with that of using DTM method to calculate the volume of the pile by GNSS-RTK for uniformly measuring the three-dimensional coordinates of the surface of the stack. It is found that the volume of the pile can meet the accuracy of the actual production.
For the calculation of stack volume with irregular shape,large scale or inconvenient closerange measurement,this paper studies the tilt photography of the stack by using a non-measured ordinary digital camera with UAV. A tilted image of the multi-view of the stack is obtained. The tilted image acquired by the drone is processed using the structure from motion and patch-based multi-view stereo( SfM-PMVS) technique. After the introduction of the ground image control point,the feature point extraction and the SIFT rough matching based on the nearest neighbor distance ratio( NNDR) algorithm are firstly applied,and the random sample consensus( RANSAC) is used to eliminate the mismatched point and then accurately obtain the basic matrix F of the image to complete image matching. The incamera parameters introduced into the calibration are used to accurately solve the essential matrix E.After the camera pose is restored,the coordinates of the sparse point cloud in the object coordinate system are calculated by the projection matrix P,and the point cloud is closely matched by the PMVS al-gorithm. Obtain a precise three-dimensional dense point cloud of the stack in the object coordinate system,conduct point cloud segmentation on the 3 D dense point cloud,and build a Delaunay triangulation after the non-stacked surface point of the stack is eliminated. The volume of the pile is calculated by the Digital Terrain Model( DTM) method. The calculation results of this method are compared with that of using DTM method to calculate the volume of the pile by GNSS-RTK for uniformly measuring the three-dimensional coordinates of the surface of the stack. It is found that the volume of the pile can meet the accuracy of the actual production.
引文
[1]王解先,侯东亚,段兵兵.三维激光扫描仪在堆积物体积计算中的应用[J].测绘通报,2013(7):54-56.WANG Jie-xian,HOU Dong-ya,DUAN Bing-bing. Application of 3D laser scanner in calculation of accumulation volume[J]. Bulletin of Surveying and Mapping,2013(7):54-56.
[2]魏占玉,Ramon,何宏林,等.基于Sf M方法的高密度点云数据生成及精度分析[J].地震地质,2015,37(2):636-648.WEI Zhan-yu,Ramon,HE Hong-lin,et al. High-density point cloud data generation and accuracy analysis based on Sf M method[J]. Seismology and Geology,2015,37(2):636-648.
[3] Glendell M,Mcshane G,Farrow L,et al. Testing the utility of structure-from-motion photogrammetry reconstructions using small unmanned aerial vehicles and ground photography to estimate the extent of upland soil erosion[J]. Earth Surface Processes&Landforms,2017,42(12):1860-1871.
[4] Yan X,Meng F,Zha H. Simplification of fan-meshes models for fast rendering of large 3D point-sampled scenes[C]//International Conference on Image Analysis and Processing. Springer Berlin Heidelberg,2005:99-106.
[5] Lorensen W E,Cline H E. Marching cubes:a high resolution 3D surface construction algorithm[J]. Acm Siggraph Computer Graphics,1987,21(4):163-169.
[6]万国伟.面向建筑物的三维点云生成、增强和重建技术研究[D].长沙:国防科学技术大学,2011.WAN Guo-wei. Research on 3D point cloud generation,enhancement and reconstruction for buildings[D].Changsha:National University of Defense Technology,2011.
[7]齐南.基于图像序列的目标三维重建技术研究[D].北京:北京工业大学,2013.QI Nan. Research on 3D reconstruction of target based on image sequence[D]. Beijing:Beijing Polytechnic University,2013.
[8]于海洋,曾春伟,马慧慧,等.基于Sf M-MVS的高山区无人机航摄数据处理[J].河南城建学院学报,2017,26(2):83-87.YU Hai-yang,ZENG Chun-wei,MA Hui-hui,et al. Aerial image processing of UAVs in high mountainous area based on Sf M-MVS[J]. Journal of Henan Institute of Urban Construction,2017,26(2):83-87.
[9]许志华,吴立新,陈绍杰,等.基于无人机影像的露天矿工程量监测分析方法[J].东北大学学报(自然科学版),2016,37(1):84-88.XU Zhi-hua,WU Li-xin,CHEN Shao-jie,et al. Monitoring and analysis method of opencast mine project based on UAV images[J]. Journal of Northeastern University(Natural Science),2016,37(1):84-88.
[10]董建伟,李海滨,孔德明,等.基于多视图立体视觉的煤场三维建模方法研究[J].燕山大学学报,2016,40(2):136-141.DONG Jian-wei,LI Hai-bin,KONG De-ming,et al.Three dimensional modeling method of Coalfield based on multiview stereo vision[J]. Journal of Yanshan University,2016,40(2):136-141.
[11]何豫航,岳俊.基于CMVS/PMVS多视角密集匹配方法的研究与实现[J].测绘地理信息,2013,38(3):20-23.HE Yu-hang,YUE Jun. Research and implementation of multiview intensive matching method based on CMVS/PMVS[J]. Surveying and Mapping Geographic Information,2013,38(3):20-23.
[12]许佳佳,张叶,张赫.基于改进Harris-SIFT算子的快速图像配准算法[J].电子测量与仪器学报,2015,29(1):48-54.XU Jia-jia,ZHANG Ye,ZHANG He. Fast image registration algorithm based on improved harris-SIFT operator[J]. Electronic Measurement&Instrumentation Journal of Engineering,2015,29(1):48-54.
[13]孙浩,王程,王润生.局部不变特征综述[J].中国图象图形学报,2011,16(2):141-151.SUN Hao,WANG Cheng,WANG Run-sheng. A review of local invariant features[J]. Chinese Journal of Image and Graphics,2011,16(2):141-151.
[14] Zhang J. Image-based 3D photography using opacity hulls[J]. Acm Transactions on Graphics,2002,21(3):427-437.
[15]陈敏.基于多尺度的图像特征提取与匹配研究[D].长沙:中南大学,2011.CHEN Min. Research on image feature extraction and matching based on multi-scale[D]. Changsha:Central South University,2011.
[16] Furukawa Y,Ponce J. Accurate,dense,and robust multiview stereopsis[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2010,32(8):1362-1376.
[17]贾毅,李振.利用空间小面元模型重构序列影像密集点云[J].武汉大学学报信息科学版,2014,39(12):1430-1434.JIA Yi,LI Zhen. Reconstructing sequence-image dense point clouds using spatial facet model[J]. Geomatics and Information Science of Wuhan University,2014,39(12):1430-1434.
[18] Ke Y,Sukthankar R. PCA-SIFT:a more distinctive representation for local image descriptors[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE Computer Society,2004:506-513.
[19] Westoby M J,Brasington J,Glasser N F,et al. Structurefrom-motion’photogrammetry:a low-cost,effective tool for geoscience applications[J]. Geomorphology,2012,179:300-314.
[20] Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision,2004,60(2):91-110.
[21]安维华,付永刚,张习文.基于图像的点云建模及其真实感绘制[J].计算机工程与应用,2010,46(20):4-8.AN Wei-hua,FU Yong-gang,ZHANG Xi-wen. Imagebased point cloud modeling and realistic rendering[J].Computer Engineering and Applications,2010,46(20):4-8.
[22]张平.基于全景相机连续影像的三维重构研究[D].合肥:安徽工业大学,2015.ZHANG Ping. Research on 3D reconstruction of continuous image based on panoramic camera[D]. Hefei:Anhui University of Technology,2015.
[23] Azevedo T C S,Tavares J M R S,Vaz M A P. 3D object reconstruction from uncalibrated images using an off-theshelf camera[J]. Lancet,2009,1(7694):348-9.
[2]魏占玉,Ramon,何宏林,等.基于Sf M方法的高密度点云数据生成及精度分析[J].地震地质,2015,37(2):636-648.WEI Zhan-yu,Ramon,HE Hong-lin,et al. High-density point cloud data generation and accuracy analysis based on Sf M method[J]. Seismology and Geology,2015,37(2):636-648.
[3] Glendell M,Mcshane G,Farrow L,et al. Testing the utility of structure-from-motion photogrammetry reconstructions using small unmanned aerial vehicles and ground photography to estimate the extent of upland soil erosion[J]. Earth Surface Processes&Landforms,2017,42(12):1860-1871.
[4] Yan X,Meng F,Zha H. Simplification of fan-meshes models for fast rendering of large 3D point-sampled scenes[C]//International Conference on Image Analysis and Processing. Springer Berlin Heidelberg,2005:99-106.
[5] Lorensen W E,Cline H E. Marching cubes:a high resolution 3D surface construction algorithm[J]. Acm Siggraph Computer Graphics,1987,21(4):163-169.
[6]万国伟.面向建筑物的三维点云生成、增强和重建技术研究[D].长沙:国防科学技术大学,2011.WAN Guo-wei. Research on 3D point cloud generation,enhancement and reconstruction for buildings[D].Changsha:National University of Defense Technology,2011.
[7]齐南.基于图像序列的目标三维重建技术研究[D].北京:北京工业大学,2013.QI Nan. Research on 3D reconstruction of target based on image sequence[D]. Beijing:Beijing Polytechnic University,2013.
[8]于海洋,曾春伟,马慧慧,等.基于Sf M-MVS的高山区无人机航摄数据处理[J].河南城建学院学报,2017,26(2):83-87.YU Hai-yang,ZENG Chun-wei,MA Hui-hui,et al. Aerial image processing of UAVs in high mountainous area based on Sf M-MVS[J]. Journal of Henan Institute of Urban Construction,2017,26(2):83-87.
[9]许志华,吴立新,陈绍杰,等.基于无人机影像的露天矿工程量监测分析方法[J].东北大学学报(自然科学版),2016,37(1):84-88.XU Zhi-hua,WU Li-xin,CHEN Shao-jie,et al. Monitoring and analysis method of opencast mine project based on UAV images[J]. Journal of Northeastern University(Natural Science),2016,37(1):84-88.
[10]董建伟,李海滨,孔德明,等.基于多视图立体视觉的煤场三维建模方法研究[J].燕山大学学报,2016,40(2):136-141.DONG Jian-wei,LI Hai-bin,KONG De-ming,et al.Three dimensional modeling method of Coalfield based on multiview stereo vision[J]. Journal of Yanshan University,2016,40(2):136-141.
[11]何豫航,岳俊.基于CMVS/PMVS多视角密集匹配方法的研究与实现[J].测绘地理信息,2013,38(3):20-23.HE Yu-hang,YUE Jun. Research and implementation of multiview intensive matching method based on CMVS/PMVS[J]. Surveying and Mapping Geographic Information,2013,38(3):20-23.
[12]许佳佳,张叶,张赫.基于改进Harris-SIFT算子的快速图像配准算法[J].电子测量与仪器学报,2015,29(1):48-54.XU Jia-jia,ZHANG Ye,ZHANG He. Fast image registration algorithm based on improved harris-SIFT operator[J]. Electronic Measurement&Instrumentation Journal of Engineering,2015,29(1):48-54.
[13]孙浩,王程,王润生.局部不变特征综述[J].中国图象图形学报,2011,16(2):141-151.SUN Hao,WANG Cheng,WANG Run-sheng. A review of local invariant features[J]. Chinese Journal of Image and Graphics,2011,16(2):141-151.
[14] Zhang J. Image-based 3D photography using opacity hulls[J]. Acm Transactions on Graphics,2002,21(3):427-437.
[15]陈敏.基于多尺度的图像特征提取与匹配研究[D].长沙:中南大学,2011.CHEN Min. Research on image feature extraction and matching based on multi-scale[D]. Changsha:Central South University,2011.
[16] Furukawa Y,Ponce J. Accurate,dense,and robust multiview stereopsis[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2010,32(8):1362-1376.
[17]贾毅,李振.利用空间小面元模型重构序列影像密集点云[J].武汉大学学报信息科学版,2014,39(12):1430-1434.JIA Yi,LI Zhen. Reconstructing sequence-image dense point clouds using spatial facet model[J]. Geomatics and Information Science of Wuhan University,2014,39(12):1430-1434.
[18] Ke Y,Sukthankar R. PCA-SIFT:a more distinctive representation for local image descriptors[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE Computer Society,2004:506-513.
[19] Westoby M J,Brasington J,Glasser N F,et al. Structurefrom-motion’photogrammetry:a low-cost,effective tool for geoscience applications[J]. Geomorphology,2012,179:300-314.
[20] Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision,2004,60(2):91-110.
[21]安维华,付永刚,张习文.基于图像的点云建模及其真实感绘制[J].计算机工程与应用,2010,46(20):4-8.AN Wei-hua,FU Yong-gang,ZHANG Xi-wen. Imagebased point cloud modeling and realistic rendering[J].Computer Engineering and Applications,2010,46(20):4-8.
[22]张平.基于全景相机连续影像的三维重构研究[D].合肥:安徽工业大学,2015.ZHANG Ping. Research on 3D reconstruction of continuous image based on panoramic camera[D]. Hefei:Anhui University of Technology,2015.
[23] Azevedo T C S,Tavares J M R S,Vaz M A P. 3D object reconstruction from uncalibrated images using an off-theshelf camera[J]. Lancet,2009,1(7694):348-9.