臺灣博碩士論文加值系統

3D印表機(3DP)為近幾年較為火熱的名詞。它用以實現許多物件的立體實體實現。隨著技術快速發展，現有3D印表機已能製作出愈發精緻的物體模型。本文欲開發一演算法--「立體拍立得」，使我們能隨時選取任一物體得其3D立體模型。
現有建立物體模型的方法大致可分為三種：電腦繪圖軟體、3D雷射掃描儀、以及影像建構。電腦繪圖軟體可以自行設計想要的物體模型；3D雷射掃描儀則可以建構出相當精準的物體模型。此二方法皆廣為使用。然而，操作電腦繪圖需要基本的三維繪圖能力，3D雷射掃描儀也不易隨身攜帶。對於想隨時隨地建構任一物體，這兩個方法就不太適合。故本文欲使用最為直覺、且操作方便的影像進行物體建模。
本文使用了一般市面手機所附的相機鏡頭，對物體進行連續兩張小角度拍攝，並在標註參考點後進行物體三維模型重建。在重建的步驟上組合了現有的基本方法如相機校正(camera calibration)、影像扭正(image rectification)、Sum of difference (SAD)，並且在全點匹配(all point matching)的搜尋範圍提出區塊配對(Block matching)的方法、匹配的計算上提出漸適性視窗尺寸(adaptive window size)與標的物計算函式(pivot cost function)方法改進。經實驗，區塊配對的方法約改善了5%的配對成功率；漸適性視窗尺寸方法約改善了17%的配對成功率；標的物計算函式約改善了6%的配對成功率；綜合所有方法則提升了約15%的配對成功率。

3D printer, which stacks an object in a 3-dimension form, is hot recent years. With the rapid development of the technique, the reconstruction of an object’s 3D model be-come more and more delicate. This work proposes an approach to realize a 3D Polaroid. Let us build any objects in a 3D form at any place in any time. In this way, a joint of 3DP would become more attractive and interesting.
So far, there are three approached to build object’s 3D model: 3D computer graphic tool, 3D laser scanner, and images’ reconstruction. We can design any models provided we want via 3D computer graphic tool. We can also get object’s model precisely by 3D computer graphic tool or we can get a precise object’s model through a 3D laser scanner. However, performing a 3D computer graphic tool usually requires a basic graphic draw-ing ability, and a 3D laser scanner is too cumbersome to become portable. To rebuild an object in a 3D form at any place in any time, this work applies images reconstruction, which is an intuitive and simple way.
In this paper, the camera embedded on the smart phone is used as the input source. Two pictures are requested to be captured in a narrow baseline along the object, and the 6 reference points should be labeled before the reconstruction. This work not only applies the well-established methods, such as camera calibration, image rectification, and sum of absolute difference method, we also improve the search range of point matching, and matching cost function with adaptive window size and pivots referencing approach in all point matching.
This work improved about 5%, 17% and 6% accuracy in block matching method, adaptive window size method and pivot cost function method, respectively. Overall, it improved about 15% accuracy for the final evaluation.

中文摘要 1
Abstract 2
誌謝 4
Content 5
Contents List of figures 7
Chapter 1 Introduction 9
1.1 Motivation 9
1.2 Outline of this thesis 10
Chapter 2 Relative work 11
2.1 Camera model 11
2.2 All point matching 14
2.3 Epipolar geometry 15
2.4 Image rectification 17
Chapter 3 Methodology 19
3.1 Data acquisition 20
3.2 Image preprocessing 21
3.3 Camera model 23
3.4 Image rectification 24
3.5 All point matching 25
3.5.1 Block matching 25
3.5.2 Adaptive window 28
3.5.3 Pivot cost function 29
3.6 Reconstruction 32
Chapter 4 Experiment ＆ Result 33
4.1 Test case 33
4.2 Ground truth acquisition 33
4.3 Experiment ＆ Result 34
Chapter 5 Conclusion ＆ Future work 47
Chapter 6 Reference 49
Appendix A 52
Appendix B 54

[1]Arne Henrichsen. 3D reconstruction and camera calibration from 2D images. PhD thesis, University of Cape Town, 2000
[2]Dyer “Volumetric scene reconstruction from multiple views. In Foundations of Image Analysis,L.S. Davis (ed.), Kluwer: Boston, MA, pp. 469–489, 2001
[3]Agarwal, S., Snavely, N., Simon, I., Seitz, S. M., ＆ Szeliski, R. Building rome in a day. InProceedings of the International Conference on Computer Vision (ICCV),2009
[4]A. Geiger, J. Ziegler, and C. Stiller. StereoScan: Dense 3d reconstruction in re-al-time. In IV, 2011
[5]R. A. Newcombe and A. J. Davison. “Live dense reconstruction with a single moving camera. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2010
[6]O. Faugeras and R. Keriven, Variational Principles, Surface Evolution, POE's, Level Set Methods, and the Stereo Problem, IEEE Transactions on 11IIIlge Processing, Vol. 7, No.3, March 1998, pp. 336-344
[7]S. M. Seitz and C. R. Dyer, “Photorealistic Scene Reconstruction by Voxel Col-oring Proc. Computer Vision and Pattern Recognition Conf., 1997, 1067-1073
[8]J. Banks, M. Bennamoun, and P. Corke. “Non-parametric techniques for fast and robust stereo matching, In Proceed- ings of IEEE TENCON, Brisbane, Australia, December 1997
[9]M. Pollefeys, R. Koch, M. Vergauwen, L. Van Gool. “Automated reconstruc-tion of 3D scenes from sequences of images. ISPRS Journal of Photogramme-try and Remote Sensing, 55 (4) (2000), pp. 251-267
[10]M Ismael, “Multi-view 3D reconstruction: a scene-based, visual hull guided, multi-stereovision framework Proceedings of the 12th European Conference on Visual Media Production Article No. 18,2015
[11]Salvi, X. Armangué, J. Batlle, “A comparative review of camera calibrating methods with accuracy evaluation Pattern Recognition, 35 (7) (2002), pp. 1617–1635
[12]Hall, E., Tio, J., McPherson, C., and Sadjadi, F. 1982. Measuring curved sur-faces for robot vision. Computer 15, 12 (Dec.), 42-54
[13]Z. Zhang, “A flexible new technique for camera calibration, IEEE Trans. Pat-tern Anal. Mach. Intell., vol. 22, no. 11, pp. 1330–1334, Nov. 2000
[14]R.I. Hartley, “Estimation of Relative Camera Positions for Uncalibrated Cam-eras, Computer Vision-ECCV ‘92, LNCS-Series vol. 588, Springer-Verlag, 1992, pp. 579–587
[15]O. Faugeras, “What can be seen in three dimensions with an uncalibrated stereo rig? In G. Sandini, editor, European Conf. Computer Vision, Santa Margherita Ligure, Italy, May 1992. Springer-Verlag
[16]A. Fusiello, “Uncalibrated Euclidean reconstruction: a review, Image and Vi-sion Computing, vol. 18, pp. 555–563, 2000
[17]MALLON, J.,AND WHELAN, P. F. “Projective rectification from the funda-mental matrix.Image Vision Computing 23, 643–650 2005.
[18]A. Fusiello and L. Irsara, “Quasi-Euclidean Uncalibrated Epipolar Rectifica-tion, ICPR, pp 1-4, 2008
[19]A. Fusiello, E. Trucco, A. Verri, “A compact algorithm for rectification of ste-reo pairs, Machine Vision Appl., 12 (1) (2000), pp. 16–22
[20]Loop, C. and Zhang, Z. 1999, “Computing rectifying homographies for stereo vision. In CVPR, Vol. I, pp. 125-131
[21]Hartley R (1999), “Theory and Practice of Projective Rectification. In: J Compute Vision 35(2): 1-16
[22]Harris, C. and Stephens, M. 1988. “A combined corner and edge detector. In Fourth Alvey Vision Conference, Manchester, UK, pp. 147-151
[23]J. Canny, “A computational approach to edge detection, IEEE Trans. Pattern Anal. Machine Intell., vol. PAMI-8, pp. 679-698, 1986