臺灣博碩士論文加值系統

本研究主要目標是利用多張二維影像重建三維物件模型後，判斷物體的
型態描述，並決定適當夾取點的位置。一般得到立體影像研究所使用的方法
大多為非接觸式系統中的雙眼視覺法，此法模擬人類雙眼,藉由視差推算物
體和攝影機之間的深度資訊用於靜態物件追蹤上，不過此方法限用於靜態物
體上，如果攝影機放置在動態的機器手臂或者是移動中的汽車上。則前人的
立體視覺法無法立即作三維重建模型。為了解決此方法,我們使用運動恢復
結構，找出連續動態拍攝的影像中，每個影像的相機關係性，並且重建出三
維模型，讓機器手臂可以判定適當的夾取點位置。
利用單眼視覺系統搭配Harris 轉角偵測方法進行影像中特徵的偵測，並依
據影像訊息求算特徵在空間中三維座標，並實現單眼視覺系統在運動恢復結
構(structure from motion)的三維模型建構。
本研究之三維建模採用三個步驟，首先將影像中的特徵點萃取出來，接著
計算出連續影像中各個物件和原始的影像差異性，並追蹤特徵點，最後利用
矩陣分解法，解出物體隱含在虛擬三維座標空間的座標，並用點雲圖呈現出
來。

This research’s object is used multi-view image rebuild 3D model, and then we use the 3D model detect the stable grasp position. Normally in stereo machine vision research, use binocular vision method is more popular. This method is likely our eyes, use disparity to calculate depth signal between camera and object.
But this method can only use on static image, if camera is putted on robot manipulator or moving car. The past method will fail to reconstruction 3D model. So in this research we will use structure from motion to find the relationship between origin cameras to another from sequence images. And we will reconstruction 3D model.
We use mono-camera get the image, and use Harris corner detector to find the keypoints. This research have 3 parts process, first extract the feature from images, and calculate the sequence images, tracking keypoints, Final we use Factorization method to get the camera and object’s relationship from 3D coordination.

摘........................................................ Ⅰ
ABSTRACT ................................................ Ⅱ
誌 謝 ................................................... Ⅲ
總目錄 .................................................. IV
圖目錄 ................................................. VII
表目錄 .................................................. IX
第一章 緒論 ............................................... 1
1.1 前言 ................................................. 1
1.2 文獻回顧 .............................................. 4
1.3 研究動機與目的 ......................................... 9
1.4 本論文之貢獻 ......................................... 10
1.5 論文架構 ............................................ 11
第二章 理論基礎 .......................................... 12
2.1 幾何圖元與轉換 ....................................... 12
2.2 鄰近點的處理 ......................................... 12
2.3 影像向量 ............................................. 13
2.4 三維座標轉換 ......................................... 15
2.5 三維對二維座標的投影 .................................. 16
2.6 相機參數關係式 ........................................ 18
2.7 運動恢復結構 ......................................... 20
第三章 影像系統建立 ....................................... 22
3.1 物體型態 ............................................ 22
3.2 二維輪廓偵測與重心 .................................... 24
3.3 輪廓編碼 ............................................. 25
3.4 夾取點判定 ........................................... 27
第四章 物體回復結構之三維模型建立 ........................... 30
4.1 影像前處理 ........................................... 30
4.2 特徵點建立 ........................................... 31
4.3 多組影像比對 ......................................... 33
4.4 矩陣分解 ............................................. 35
4.5 三維模型建構 ......................................... 38
第五章 實驗結果與討論 ..................................... 39
5.1 實驗設備 ............................................. 39
5.2 物體型態辨識系統操作流程 ............................... 42
5.3 平面物體二維夾取點分析 ................................. 42
5.4 三維模型之重建 ........................................ 47
第六章 結論與未來展望 ..................................... 53
參考文獻 ................................................. 54

[1] Vittorio Belotti, Manjula U. Hemapala, Rinaldo C. Michelini, and Roberto P. Razzoli, “Lean Robotics For Humanitarian Mine Sweeping,” International Conference Advanced Robotics, pp. 1-6, 2009
[2] Fu Wei Cao Jie Bian Qingqing and Yang Jin,” Image processing and control strategy of tele-operating system of a Bomb- Disposed Robot”, International Conference on Embedded Software and Systems Symposia , pp. 453-459,2008.
[3] David L. Bowers and Ron LumiaProfumo, “Manipulation of Unmodeled Objects Using Intelligent Grasping Schemes,” IEEE Transcations on Fuzzy Systems, Vol. 11,pp. 320-329, 2003.
[4] S. Cortez, J.C.P. Claro, J.L. Alves, “3D Reconstruction of a Spinal Motion Segment from 2D Medical Images: Objective Quantification of the Geometric Accuracy of the FE Mesh Generation Procedure,” IEEE conference in Portuguese Meeting in Bioengineering, pp. 1-6, 2013
[5] Wang Hongjian, and Pu Xiaoming, “3D Medical CT Images Reconstruction based on VTK and Visual C++,” International Conference on Bioinformatics and Biomedical Engineering , pp.1-4 , 2009
[6] Fr´ed´eric Pighin, Richard Szeliski,and David H. Salesin, “Modeling and Animating Realistic Faces from Images,” International Journal Computer Vision, Vo1.9, No 2, 2002.
[7] Huang Ying Yuan, “3D Object Model Recovery from 2D Images Utilizing Corner Detection,”Taiwan, R.O.C.,2010.
[8] Pedro J. Sanz, Angel P. del Pobil, and Joses M. Iniesta, “Real-Time Grasping of Unknown Objects Based on Computer Vision System”, International Conference on Advanced Robotics, pp. 319-324,1997.
[9] Eris Chinellato, Antonio Morales, Robert B. Fisher, and Angel P. del Pobil, “Visual Quality Measures for Characterizing Planar Robot Grasps,” IEEE Trans. on syst., Man, and cybernetics.-part C, VOL. 35, No.1, pp.30-41 , 2005.
[10] Eris Chinellato, Member, IEEE, Antonio Morales, Member, IEEE, Robert B. Fisher and Angel P. del Pobil , “Visual Quality Measures for Characterizing Planar Robot Grasps,” IEEE Pattern Trans. on syst., Man, and cybernetics, pp. 1-12, 2007
[11] Mila Popovi¢ , Dirk Kraft, Leon Bodenhagena, Emre Ba³eski, Nicolas Pugeault, Danica Kragi , Tamim Asfour, and Norbert Krüger ,” A strategy for grasping unknown objects based on co-planarity and colour information,” Robtics and Autonomous Systems, Vol.58, pp.551-565 , 2010.
[12] Vincenzo Lippiello, Fabio Ruggiero, Member, IEEE, Bruno Siciliano, and Luigi Villani, “Visual Grasp Planning for Unknown Objects Using a Multifingered Robotic Hand,” IEEE Tran. On Metronics., Vol. 6, pp. 1-10, 2010.
[13] Jeff Fortuna, and Aleix M. Martinez, “Rigid Structure from Motion from a Blind Source Separation, ” Internation Journal Computer Vision , Vol. 88,pp. 404-424, 2010.
[14] S. Suzuki and K. Abe, “Topological Structural Analysis of Digitized Binary Images by Border,” CVGIP, vol.30, pp.32-46,1985
[15] Carlo Tomasi, and Takeo Kanade, “Shape and Motion from Image Streams under Orthography: a Factorization Method,” International Journal of Computer Vision, Vol.9, pp.137-154, 1992
[16] Lowe D.G., “Distinctive Image Features from Scale-Invariant Keypoints,” International Journal of Computer Vision, Vol.60, pp.91-110, 2004