臺灣博碩士論文加值系統

本篇論文運用多個環場影像並採用立體求形(Stereo Matching)方法，取得場景的3D資訊，分兩個方向進行，多基線立體求形( Multiple-Baseline Stereo)和體素著色(Voxel Coloring)。
多基線立體求形運用SSSD (Sum of Sum of Square Difference)比對方式找到對應點，即可得到場景的三度空間座標，經過建構即可產生3D場景；另外我們利用分析SSSD值，使得我們可以找到發生遮蔽現象和平滑的區域，先將這些區域捨棄掉，最後再透過線性內插去產生這些區域的三度空間座標，避免去找到錯誤的點。
體素著色(Voxel Coloring)是最近比較熱門的立體求形方法，它可以針對上述方法中較難處理的遮蔽現象加以改進，此方法將物件由一個盒形範圍 (Bounding Box)包圍，此盒形範圍由大小相同的體素組合成，透過體素一致性(Voxel Consistency)的檢查決定此體素存不存在，最後得到物件的模型。
根據體素著色(Voxel Coloring)我們提出了智慧型自動取樣方法。這個智慧型自動取樣的目的，是期望用最少的攝影機還原整個場景，一般可以延伸到自動化機器人系統，在房間形狀不明的情況下，運用機器人來還原房間的場景。

This study aims at obtaining 3D structure of scene by acquiring multiple panoramas and applying stereo matching methods. Two stereo matching methods have been considered in this thesis ; Multiple-baseline stereo and voxel coloring.
Multi-baseline stereo method utilizes SSSD(sum of sum of square difference) to find corresponding points to estimate 3D space coordinates. Besides, after analyzing SSSD, we can detect the occluded and textureless regions . Then, 3D space coordinates of these regions are refilled through linear interpolation and outliers are suppressed.
Voxel Coloring is a popular stereo matching method in that space is divided into many small cubes , called voxels. By examining the color consistency of each voxel , we can determine whether a voxel exist or should be removed . Thus , the 3D structure of the scene can be estimated.
Adding extra cameras are necessary since there exist some invisible regions where can’t be seen by existing cameras. This study proposed a smart sampling method using voxel coloring method. The purpose is to restore the most regions of scene by using the minimum number of cameras. In practice, it can be apply to automatic robot system to restore 3D scene structure without knowing the layout of the space in advance.

第一章緒論……………………………………………………………………..1
1.1概說…………………………………………………………………1
1.2目標與系統概觀……………………………………………………3
第二章相關研究………………………………………………………………..4
2.1 多基線立體求形(Multiple Baseline Stereo)…………………4
2.2 體素著色(Voxel Coloring)…………………………………...6
第三章環場影像的建構………………………………………………………..9
3.1 虛擬環場影像的建構……………………………………………..9
3.2 真實環場影像的建構……………………………………………..10
第四章使用多基線立體求形(Multiple-Baseline Stereo)方法建構3D場景…18
4.1 投影和反投影……………………………………………………..20
4.2多基線立體求形(Multiple-Baseline Stereo)相似度比對…..……...21
4.3 運用權重整合SSSD………………………………………………22
4.4分析SSSD值……………………………………………………….24
4.5攝影機校正不準或影像扭曲不正確時之處理方式………………25
第五章體素著色(Voxel Coloring)和智慧型自動取樣方法……………….…28
5.1使用體素著色還原場景…………………………………………...28
5.2 尋訪順序…………………………………………………………..32
5.3 可見度的判斷……………………………………………………..33
5.4智慧型自動取樣……………………………………………………36
第六章實驗結果……………………………………………………………….39
6.1多基線立體求形(Multiple Baseline Stereo)實驗結果…………….39
6.2體素著色(Voxel Coloring)實驗結果……………………………….46
第七章結論和未來研究建議…………………………………………………..55
7.1結論…………………………………………………………………55
7.2未來研究建議………………………………………………………55

[1]Bunschoten, R. and Krose, B., “Robust Scene Reconstruction from an Omnidirectional Vision System”, IEEE Transactions on Robotics and Automation, Vol. 19, Issue 2, pp.351 — 357, Apr. 2003.
[2]Bunschoten, R. and Krose, B., “3-D Scene Reconstruction from Cylindrical Panoramic Images”, Robotics and Autonomous Systems (special issue), Vol. 41(2/3), pp.111-118 , Nov. 2002.
[3]Eisert, P., Steinbach, E. and Girod, B., “Multi-Hypothesis, Volumetric Reconstruction of 3D Objects from Multiple Calibrated Camera Views”, ICASSP’99 ,pp. 3509-3512, Mar. 1999.
[4]Jeon, J., Kim, K., Kin, C. and Ho, Y.S., “A Robust Stereo-Matching Algorithm Using Multiple-Baseline Cameras”, IEEE Pacific Rim Conference on Computers and Signal Processing, Vol.1, pp.263 — 266, Aug. 2001.
[5]Jia, B., Zhang, Y.J. and Lin X.G., “Stereo Matching Using Both Orthogonal and Multiple Image Pairs”, IEEE International Conference on Acoustics, Speech, and Signal Processing, Vol.6, pp.2139— 2142, Jun. 2000.
[6]Jain, R., Kasturi, R.C. and Schunck, B.G., “Machine Vision”, McGraw-Hill, 1995.
[7]Ku, J.S., Lee, K.M. and Lee, S.U., “Multi-Image Matching for a General Motion Stereo Camera Model”, ICIP 98 on Image Processing, Vol. 2, pp. 608 — 612, Oct. 1998.
[8]Kanade, T., “Development of a Video-Rate Stereo Machine”, 1995 IEEE/RSJ International Conference on Intelligent Robots and Systems 95, Human Robot Interaction and Cooperative Robots, Proceedings, Vol.3, pp.95 — 100, Aug. 1995.
[9]Kang, S.B. and Szeliski, R., “3-D Scene Data Recovery using Omnidirectional Multibaseline Stereo”, IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp.364—370, June 1996.
[10]Klein, K. and Sequeira, V., “The View-Cube: An Efficient Method of View Planning for 3D Modeling from Range Data”, Proceedings of the 5th IEEE Workshop on Applications of Computer Vision (WACV'2000) , pp.186-191, Dec. 2000.
[11]Li, Y., Lin, S., and Lu, Hu., “Multibaseline Stereo in the Presence of Specular Reflections”, 16th International Conference on Pattern Recognition,Vol.3, pp.573 — 576, Aug. 2002.
[12]Moezzi, S., Tai, L.C. and Gerard, P., “Virtual View Generation for 3D Digital Video”, IEEE Multimedia, Vol. 4, Issue 1, pp.18 — 26, Mar. 1997.
[13]Okutomi, M. and Kanade, T., “A Multiple-Baseline Stereo”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.15, Issue 4, pp. 353 — 363, Apr. 1993.
[14]Park, J.I. and Inoue, S., “Acquisition of Sharp Depth Map from Multiple Cameras”, Signal Processing: Image Communication, Vol.14, pp.7-19, Nov. 1998.
[15]Slabaugh, G.G., Culbertson, W.B., Malzbender, T. and Schafer, R.W., “A Survey of Methods for Volumetric Scene Reconstruction from Photographs”, International Workshop on Volume Graphics, Jun. 2001.
[16]Seitz, S.M. and Dyer, C.R., “Photorealistic Scene Reconstruction by Voxel Coloring”, IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp.1067 — 1073, Jun. 1997.
[17]Seitz, S.M. and Kutulakos, K.N., “A Theory of Shape by Space Carving” The Proceedings of the Seventh IEEE International Conference on Computer Vision, Vol.1, pp.307 — 314, Sep. 1999.
[18]Steinbach, E., Girod, B., Eisert, P. and Betz A., “3-D Object Reconstruction Using Spatially Extended Voxels and Multi-Hypothesis Voxel Coloring”, 15th International Conference on Pattern Recognition, Vol.1, pp.774 — 777, Sep. 2000.
[19]Umbaugh, S.E , “Computer Vision and Image Processing:A Practical Approach Using CVIPtools”, Prentice Hall ,1998.
[20]Yang, R., Pollefeys, M. and Welch, G, “Dealing with Texureless Regions and Specular Highlights — A Progressive Space Carving Scheme Using a Novel Photo-consistency Measure”, Ninth IEEE International Conference on Computer Vision, pp.576 — 584, Oct. 2003.
[21]Zhu, Z., “Omnidirectional Stereo Vision”, The 10th IEEE International Conference on Advanced Robotics, Aug. 2001.