臺灣博碩士論文加值系統

以照片為來源的三維建模系統，關鍵在於正確地取得相機的內外部參數，傳統的做法是在欲重建模型邊擺上棋盤式的校正板 (Chessboard)，透過校正板去做相機參數估測，得到相機位置後利用Visual Hull求出三維模型，但這種做法有太多限制，除了需事先備好校正板，要架設整個環境亦需一番苦工。本文提出以特徵點為基礎的自動建模系統，利用尺度不變特徵轉換(SIFT)在輸入的影像上尋找角點並做角點對應，角點對應可以用來估計相機參數。得到相機參數後，將特徵點投影回三維空間，再透過柏松(Poisson)平面重建演算法，把點雲表面還原。在影像的取得上，我們改用手持物體的方式，不僅增加便利性也可以獲得更多死角資訊，提升模型精確度。但由於手持會產生過多雜訊，影響模型後模型之幾何構造，為解決這個問題，我們設計了一套針對手部的去雜訊演算法。

To estimate the precise intrinsic and extrinsic parameter of the camera is the key point for the design of 3D reconstruction systems. Traditionally, we should calibrate the camera parameters with a chessboard, followed by visual hull to generate a closed 3D model. Nevertheless, there are some limitations, which need to prepare a chessboard, and take much energy to construct the device. In the proposed method, a feature point based system is used to find the camera matrix. Then, the sparse point cloud is reconstructed while estimating the camera matrix. Finally, Poisson surface reconstruction is applied to create the surface. In required images, we allow the user to hold the object in front of a fixed camera. This method is not only more convenient but can record more information in hidden area. But, the reconstructed point cloud is noisy because of the inference of user’s hands, so a noise removal step is designed in our system to solve this problem.

List of Figures vi
1 Introduction 1
1.1 Motivation 1
1.2 Related Research 2
1.3 Problem Statement and Contribution 4
1.4 Thesis Organization 6
2 Basic Concepts of Reconstruction 8
2.1 Camera model 8
2.1.1 Extrinsic parameters 9
2.1.2 Intrinsic parameters 11
2.2 Epipolar Geometry 14
2.2.1 Fundamental matrix and essential matrix 15
2.3 RANSAC 18
2.4 Scale invariant feature transformation 21
2.4.1 Scale space 22
2.4.2 Laplacian of Gaussian approximation 23
2.4.3 Local extrema detection 24
2.4.4 Orientation assignment 26
2.4.5 Generate SIFT feature descriptor 27
2.4.6 SIFT feature matching 28
3 An Automatic 3D Reconstruction System 30
3.1 Feature extraction 31
3.1.1 SIFT feature detector 31
3.1.2 Bad matching removal and SIFT flow 32
3.2 Structure from Motion 36
3.2.1 Initial structure 36
3.2.1 Updating the structure 40
3.3 Noise removal 43
4 Experimental Results 48
4.1 Advantages of handheld object 49
4.2 Performance of noise removal 51
4.3 Operating time and quality analisis 52
4.4 Applications of 3D reconstruction 53
4.5 Other model 55
5 Conclusions 59
References 60

[1]Ze-tao Jiang , Bi-na Zheng, Min Wu, Wen-huan Wu. A Fully Automatic 3D Reconstruction Method Based on Images. World Congress on Computer Science and Information Engineering, 2009.
[2]Noah Snavely, Steven M. Seitz, Richard Szeliski. Photo Tourism: Exploring Photo Collections in 3D. ACM Transactions on Graphics (Proceedings of SIGGRAPH ),2006.
[3Noah Snavely, Steven M. Seitz, Richard Szeliski. Modeling the World from Internet Photo Collections. International Journal of Computer Vision, 2007.
[4]Sameer Agarwal, Noah Snavely, Ian Simon, Steven M. Seitz, Richard Szeliski. Building Rome in a Day, Computer Vision, IEEE 12th International Conference, 2009
[5]David G. Lowe, Distinctive image features from scale-invariant keypoints, International Journal of Computer Vision, 60, 2 (2004), pp. 91-110.
[6]David G. Lowe, Object recognition from local scale-invariant features, International Conference on Computer Vision, Corfu, Greece (September 1999), pp. 1150-1157.
[7]David G. Lowe, Local feature view clustering for 3D object recognition, IEEE Conference on Computer Vision and Pattern Recognition, Kauai, Hawaii (December 2001), pp. 682-688.
[8]Changchang Wu, SiftGPU: A GPU implementation of Scale Invaraint Feature Transform (SIFT), http://cs.unc.edu/~ccwu/siftgpu, 2007.
[9]Changchang Wu, Sameer Agarwal, Brian Curless, and Steven M. Seitz, Multicore Bundle Adjustment, CVPR 2011.
[10] K. Forbes, A. Voight, and N. Bodika, An inexpensive automatic and accurate camera calibration method, in Proceedings of Thirteenth Annual Symposium of the Pattern Recognition Association of South Africa, Langebaan, South Africa, Nov. 2002, pp.100-106.
[11]V. Douskos, I. Kalisperakis, and G. Karras Automatic calibration of digital cameras using planar chess-board patterns, in Proceeding of the 8th Conference on Optical 3-D Measurement Techniques, ETH Zurich, Switzerland, July 2007, pp. 132-140.
[12]Yanli Wan, Zhenjiang Miao, 3D Scene Reconstruction Based on Uncalibrated Image Sequences, International Conference on Digital Image Processing.
[13]M. Kazhdan, M. Bolitho and H. Hoppe, Poisson surface reconstruction, Geometry Processing Conference, Vol. 256, pp. 61-70, 2006.
[14]TZUR, Y., TAL, A., FlexiStickers - Photogrammetric Texture Mapping using Casual Images, SIGGRAPH 2009, ACM Transactions on Graphics, Volume 28, Issue 3, August ,2009
[15]David Niste´r, An Efficient Solution to the Five-Point Relative Pose Problem, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 26, NO. 6, JUNE 2004.
[16]Michel Sarkis, Klaus Diepold, Knut H¨uper, A FAST AND ROBUST SOLUTION TO THE FIVE-POINT RELATIVE POSE PROBLEM USING GAUSS-NEWTON OPTIMIZATION ON A MANIFOLD, Technische Universit¨ at M¨unchen Institute for Data Processing Munich, Germany,2007
[17]Batra, D., Nabbe, B., Hebert, M. An Alternative Formulation for Five Point Relative Pose Problem, Motion and Video Computing, 2007.
[18]Hongdong Li, Hartley, R., Five-Point Motion Estimation Made Easy, IEEE 18th International Conference on Pattern Recognition, 2006.

[19]Zuzana Kukelova, Martin Bujnak and Tomas Pajdla, Polynomial Eigenvalue Solutions to the 5-pt and 6-pt Relative Pose Problems, BMVC 2008.
[20]Shengjun Liu, Kwan-Chung Chan, and Charlie C.L. Wang. Iterative Consolidation of Unorganized Point Clouds, IEEE COMPUTER GRAPHICS AND APPLICATIONS, 2012.
[21]Hui Huang, Dan Li, Hao Zhang, Uri Ascher, Daniel Cohen-Or. Consolidation of Unorganized Point Clouds for Surface Reconstruction, ACM Transactions on Graphics (Proceeding of SIGGRAPH Asia), 2009
[22]Yasutaka Furukawa, Brian Curless, Steven M. Seitz, Uri Ascher, Richard Szeliski. Towards Internet-scale Multi-view Stereo IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2010
[23]Yasutaka Furukawa, Brian Curless, Steven M. Seitz, Uri Ascher, Richard Szeliski. Towards Internet-scale Multi-view Stereo IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2010
[24]Chanop Silpa-Anan, Richard Hartley, Optimised KD-trees for fast image descriptor matching, IEEE Conference on Computer Vision and Pattern Recognition, 2008.
[25]Jeffrey S. Beis, David G. Lowe, Shape Indexing Using Approximate Nearest-Neighbour Search in High-Dimensional Spaces, IEEE Conference on Computer Vision and Pattern Recognition, 1997.
[26]Sunil Arya, David M. Mount, Nathan S. Netanyahu, Ruth Silverman, Angela Y. Wu An Optimal Algorithm for Approximate Nearest Neighbor Searching in Fixed Dimensions, Journal of the ACM, Volume 45 Issue 6, Nov. 1998.
[27]Mikolajczyk, K. Detection of local features invariant to affine transformations, Ph.D. thesis, Institut National Polytechnique de Grenoble, France, 2002.