臺灣博碩士論文加值系統

因為現有技術是假設所搜尋的3D模型是剛體，因此有關節相連的3D模型搜尋結果便不是那麼完美。因此我們提出了一個新的方法，讓使用者可以從資料庫中搜尋出有關節相連的3D模型。

在此篇paper中，我們提出了一個方法，可同時使用Global 和Local資訊來搜尋3D模型。Global資訊即為原本的2D 圖像(3D模型的投影圖 或是 深度照相機的某個frame)。因為有關節相連的3D 模型可能會有不同的姿勢(不同的轉角、不同的手部動作、不同腳步動作等)，因此我們需要Local資訊。Local資訊即為Global資訊的Local形狀。舉個例來說，一張人型的圖片為Global資訊，其中手、腳、身體即為Local資訊。

利用Local資訊我們可以找到更多不同姿勢的3D model，即是可以增加辨別率。為了解決不同姿勢和noise的問題，我們提出了一個新的feature set可以更好的處理這些情況以及四腳動物的搜尋。

我們方法的優點為我們不需要將模型轉到一個特定的角度才能做搜尋。更棒的是，我們平均的搜尋時間小於1秒，幾近即時的搜尋系統。

Existing methods can barely get satisfactory results since they usually assume the objects are rigid. Therefore we present a novel approach that we can retrieve articulate 3D models from database.

In this paper, we propose a novel method to retrieve models by using both global and local information. The global information is the original 2D views (projected views of a 3D model or frames of a depth camera). Since the articulate 3D models may have various pose (various orientations and diverse motions of body parts or limbs), we still need local information. The local information is the local shapes of the 2D views. Take a human image for example, the hands, lags, and the body are the local information of the this image.

By using the local information, the recognition rate is increasing, because we can retrieve much more models with different pose. Meanwhile, we also define a feature set which is robust against various pose and noise. By using our feature set, we can deal with some special case such as 4-lags animal models.

The advantage of our method is that we do not need a well-aligned model pose. Best of all, since the cost average of the cost time of each inputs is under 1 seconds, the online retrieve system is almost real-time.

摘要 I
Abstract II
Acknowledgement III
Table of Contents IV
List of Figures V
List of Table VI


Chapter 1. - 1 -
1.1 Motivation - 1 -
1.2 Overview - 3 -

Chapter 2. - 5 -
2.1 Applications of RGB-D camera - 5 -
2.2 Retrieval system - 6 -
2.3 Skeleton extraction method - 7 -

Chapter 3. - 9 -
3.1 Extraction of Views - 10 -
3.2 Extraction of local information - 10 -
3.3 Extraction of features - 11 -
3.4 Databases - 17 -

Chapter 4. - 19 -
4.1 Get inputs and extract features - 19 -
4.2 Matching method - 21 -
4.3 Distance rule - 23 -

Chapter 5. - 24 -
5.1 Retrieval system - 24 -
5.2 Performance Evaluation - 25 -

Chapter 6. - 29 -

References - 30 -

[1] T. Igarashi , S. Matsuoka , H. Tanaka, “Teddy: A sketching interface for 3d freeform design”, ACM SIGGRAPH 1999, Proceedings of the 26th annual conference on Computer graphics and interactive techniques, pp. 409-416
[2] K. Lai, L. Bo, X. Ren, and D. Fox, “A Large-Scale Hierarchical Multi-View RGB-D Object Dataset”, IEEE International Conference on Robotics and Automation 2011, pp. 1817-1824
[3] L. Bo, X. Ren, and D. Fox, ”Depth Kernel Descriptors for Object Recognition”, IEEE Intelligent Robots and System 2011, pp. 821-826
[4] W. Chang, M. Zwicker, “Global Registration of Dynamic Range Scans for Articulated Model Reconstruction”, ACM Transactions on Graphics 2011, Vol. 30, No.26
[5] D. Chen, X. Tian, Y. Shen and M. Ouhyoung, "On Visual Similarity Based 3D Model Retrieval", Computer Graphics Forum (EUROGRAPHICS 2003), Vol. 22, No. 3, pp. 223-232
[6] T. Shao, W. Xu, K. Yin, J. Wang, K. Zhou, and B. Guo, "Discriminative sketch-based 3D Model Retrieval via Robust Shape Matching", Computer Graphics Forum (PG 2011), Vol.30, N0. 7, pp. 2011-2020
[7] P. Daras, A. Axenopoulos, “A 3D Shape Retrieval Framework Supporting Multimodal Queries”, SPRINGER, International Journal of Computer Vision 2010, pp.229-247
[8] N. Canterakis, ” 3D Zernike moments and Zernike affine invariants for 3D image analysis and recognition”, Scandinavian conference on image analysis 1999.
[9] M. Novotni, R. Klein, “3D zernike descriptors for content based shape retrieval”, ACM symposium on Solid modeling and applications 2003, pp.216-225
[10] P. Yap, R. Paramesran, S. Ong, “Image analysis by Krawtchouk moments”, IEEE Transactions on Image Processing 2003, Vol.12, No. 2003, pp.1367-1377
[11] Y. Ke and R. Sukthankar, "PCA-SIFT: A More Distinctive Representation for Local Image Descriptors", IEEE Computer Vision and Pattern Recognition 2004, pp.506-513.
[12] W. Jiang, K. Xu, Z. Cheng, R.. Martin, G. Dang, “Curve Skeleton Extraction by Coupled Graph Contraction and Surface Clustering”, Computation Visual Media Conference 2012, pp. 178-185
[13] D. Zhang, and G. Lu, “Shape-based image retrieval using generic Fourier Descriptor”, . ELSEVIER Signal Processing: Image Communication 2002, Vol.17, No.10, pp.825–848
[14] Y. Gao, J. Tang,S. Yan, Q. Dai,N. Zhang, and T. Chua, ” Camera Constraint-Free View-Based 3-D Object Retrieval”, IEEE TRANSACTIONS ON IMAGE PROCESSING 2012, Vol. 21, N0. 4, pp2269-2281
[15] S. Izadi , D. Kim , O. Hilliges , D. Molyneaux , R. Newcombe , P. Kohli , J. Shotton , S. Hodges , D. Freeman , A. Davison , and A. Fitzgibbon, “Kinectfusion: real-time 3D reconstruction and interaction using a moving depth camera ”, ACM Symposion on User Interface Software and Technology 2011, pp.559-568
[16] J. Cao, A. liasacchi, M. Olson, H. Zhang ,and Z. Su, “Point Cloud Skeletons via Laplacian Based Contraction”, IEEE Shape Modeling International Conference 2010, pp.187-197
[17]ASUS Xtion Pro http://www.asus.com/Multimedia/Xtion_PRO/
[18]Microsoft Kinect http://www.xbox.com/zh-TW/Kinect
[19]PrimeSense http://www.primesense.com/
[20]P. Merrell, A. Akbarzadeh, L. Wang ,P. Mordohai, ” real-time visibility-based fusion of depth maps”, IEEE International Conference on Computer Vision 2007, pp.1-8
[21] P. Henry, M. Krainin, E. Herbst, X. Ren, D. Fox, ” RGB-D Mapping: Using Depth Cameras for Dense 3D Modeling of Indoor Environments”, International Symposium on Experimental Robotics 2010
[22]H. Sundar, D. Silver, N.Gagvani, and S. Dickinson, “skeleton based shape matching and retrieval”,ACM Shape Modeling International 2003, pp.130-140
[23]D. Lowe, “Object Recognition from Local Scale-Invariant Features”, IEEE Computer Vision 1999, Vol. 2, pp. 1150-1157