[1]姜伟,宋振东,魏世恒.稀疏图像空间全周场景图像化和三维重建[J].南京理工大学学报(自然科学版),2012,36(05):854.
 JIANG Wei,SONG Zhen-dong,WEI Shi-heng.Panoramic Imaging and 3D Reconstruction Based onSparse Image Volume[J].Journal of Nanjing University of Science and Technology,2012,36(05):854.
点击复制

稀疏图像空间全周场景图像化和三维重建
分享到:

《南京理工大学学报》(自然科学版)[ISSN:1005-9830/CN:32-1397/N]

卷:
36卷
期数:
2012年05期
页码:
854
栏目:
出版日期:
2012-10-31

文章信息/Info

Title:
Panoramic Imaging and 3D Reconstruction Based onSparse Image Volume
作者:
姜伟1宋振东2魏世恒1
1. 浙江大学工业控制技术国家重点实验室,浙江杭州310027;2. 黑龙江大学信息科学与技术学院,黑龙江哈尔滨150080
Author(s):
JIANG Wei1SONG Zhen-dong 2WEI Shi-heng 1
1. State Key Laboratory of Industrial Control Technology,Zhejiang University,Hangzhou 310027,China;2. College of Information Sciences and Technology,Heilongjiang University,Haerbin 150080,China
关键词:
全周三维重构全景图稀疏图像空间多基线算法视野选择
Keywords:
panoramic 3D reconstructionpanoramasparse image volumemulti-baseline algorithmscamera selection
分类号:
TP391. 4;TP242. 6
摘要:
针对全周场景图像化和三维重建问题,该文通过匀速旋转一台非中心配置相机收集全周场景的时间序列图像,生成稀疏图像空间;在分析图像空间外极约束关系的基础上,导出图像空间中特征点轨迹形状和目标点深度信息间的关系,并通过具有视野选择功能的多基线算法抽出稀疏轨迹,实现了空间点的深度信息提取;最后通过插值生成场景的全景图。稀疏图像空间的应用提高了图像化速度,多基线算法提高了深度信息提取的鲁棒性,相机选择有效地解决了立体视觉中的“深度遮挡"问题。实验结果显示,该方法能够提高摄像速度45 倍、降低均方根误差95%以上,高可靠性地实现了全周场景图像化和三维重构。
Abstract:
A new approach for obtaining a dense panorama image and 3D depth information aboutpanoramic(360°) environments is presented here. A sparse spatio-temporal volume is obtained byrotating a noncentral image acquisition rig to collect the panorama. By using the multi-baseline stereotechnique on the sparse spatio-temporal volume, the panoramic depth map is estimated and thepanorama image with the same spatial resolution as the original regular images is generated. The newapproach produces better results than previous approaches and shortens acquisition time of regularimages by enlarging the rotational angular interval. Experimental results show that this approach canincrease imaging speed 45 times,decrease root-mean-square error( RMSE) over 95%,and producehigh quality panorama image and panoramic 3D reconstruction.

参考文献/References:

[1]Baker S, Nayar S K. A theory of single-viewpointcatadioptric image formation[ J]. International Journalof Computer Vision,1999,35(2):1-99.
[2] Yi S Y,Ahuja N. An omnidirectional stereo vision systemusing a single camera[A]. Proceedings of InternationalConference on Pattern Recognition [ C]. Hong Kong,China:IEEE Press,2006:29-38.
[3] Cheng S E. Quicktime VR—An image-based approachto virtual environment navigation[ A]. Proceedings ofACM Siggraph [ C ]. Los Angeles, CA, US: ACMSiggraph,1995:29-38.
[4] Szeliski R. Video mosaics for virtual environment[ J].IEEE Computer Graphics and Applications,1996,16(3):22-30.
[5] Kang S B,Szeliski R. 3-D scene data recovery usingomnidirectional multibaseline stereo[ J]. InternationalJournal of Computer Vision,1997,25(2):167-183.
[6] Alex R A,Dani L,Shmuel P. Dynamosaicing:mosaicingof dynamic scenes[ J]. IEEE Transactions on PatternAnalysis and Machine Intelligence, 2007, 29 ( 10):1789-1801.
[7] Jiang W,Sugimoto S,Okutomi M. Panoramic 3D recon-struction using rotational stereo camera with simpleepipolar constraints [ A ]. Proceedings of IEEEComputer Society Conference on Computer Vision andPattern Recognition[C]. New York,USA:IEEE Press,2006:371-378.
[8] Li Y,Shum H Y,Tang C K. A stereo reconstruction frommulti-perspective panoramas[J]. IEEE Transactions onPattern Analysis and Machine Intelligence, 2004, 26(1):45-62.
[9] Li R,Yan L,Di L,et al. A new ground-based stereopanoramic scanning system [ A ]. Proceedings ofInternational Archives of the Photogrammetry,RemoteSensing and Spatial Information Sciences[C]. Beijing,China:ISPRS,2008:723-727.
[10] Peleg S, Ben-Ezra M, Y Pritch. Omnistereo: panoramicstereo imaging[J]. IEEE Transactions on Pattern Analysisand Machine Intelligence,2001,23(3):279-290.
[11] Okutomi M, Katayama Y, Oka S. A simple stereoalgorithm to recover precise object boundaries andsmooth surfaces[J]. International Journal of ComputerVision,2002,47(1-3):261-273.
[12] Makadial A, Geyer C, Daniilidis K. Correspondence-free structure from motion[J]. International Journal ofComputer Vision,2007,72(3):311-327.
[13] Okutomi M,Kanade T. A multiple-baseline stereo[J].IEEE Transactions on Pattern Analysis and MachineIntelligence,1993,15(4):353-363.

备注/Memo

备注/Memo:
收稿日期:2010-07-28修回日期:2012-03-10基金项目:浙江省自然科学基金(Y1110067);浙江省钱江人才计划项目(2010R10003)作者简介:姜伟(1969-),男,副教授,主要研究方向:机器视觉、模式识别、智能系统,E-mail:jiangwei@ iipc. zju.edu. cn;通讯作者:宋振东(1964-),男,副教授,主要研究方向:数值计算、模式识别,E-mail:songzd@hlju. edu. cn。
更新日期/Last Update: 2012-11-26