臺灣博碩士論文加值系統

為了因應與日俱增的多媒體資料，人們不斷探討如何有效對多媒體資料進行管理與檢索。過去相關研究大多著重在二維影像及三維模型上，但隨著日益增加的RGB-D 影像資料，檢索方法仍顯得有所不足，因而開發相關檢索技術以利於搜尋已是刻不容緩的工作。本研究即以Kinect 感測裝置所取得的RGB-D 影像為來源，探討如何從點雲資料萃取同時具備色彩與幾何之特徵，並以此為基礎設計新的物件檢索方法。為了展現執行成效，實際建立測試平台以驗證其實際成效。

Nowadays, modern multimedia are widely used in daily life. In order to response the requirements, people continuously explore how to effectively manage and retrieve multimedia data. Most previous studies focused on 2D images and 3D models. However, along with the increasing RGB-D image data, the related retrieval technologies are still inadequate. Therefore, it is an urgent task for development of related retrieval algorithms. This research uses RGB-D image data obtained by Kinect sensing device as the input source, and studies how to extract both color and geometry features from the point cloud data and design a new 3D object retrieval system. By exhibiting the results, a testing platform is established to verify actual effectiveness.

銘謝 ......................... I
摘要 ......................... II
Abstract ..................... III
目錄 ......................... IV
圖目錄 ....................... VII
表目錄 ....................... IX
一、緒論 ..................... 1
1.1 研究動機 ................. 1
1.2 研究目的 ................. 2
二、相關文獻回顧 .............. 3
2.1 深度攝影機 ................ 3
2.1.1 三維掃描儀 .............. 3
2.1.2 深度偵測原理 ............ 4
2.1.3 Kinect 感測裝置 ......... 6
2.2 點雲資料 .................. 7
2.3 資料檢索方法 .............. 8
2.3.1 二維影像 ................ 8
2.3.2 三維幾何模型 ............ 11
2.3.3 RGB-D 影像 .............. 12
三、研究方法 ................... 14
3.1 取得RGB-D 影像 ............. 14
3.2 前處理 ..................... 15
3.2.1 視角校正 ................. 16
3.2.2 物件分割 ................. 17
3.2.3 點雲轉換 ................. 17
3.3 特徵萃取 ................... 18
3.3.1 D2 特徵萃取 .............. 19
3.3.2 D3 特徵萃取 .............. 21
3.3.3 A3 特徵萃取 .............. 21
3.4 特徵正規化 ................. 21
3.5 特徵比對 ................... 22
四、實驗方法與討論 .............. 24
4.1 使用者介面 ................. 27
4.2 特徵萃取數實驗 .............. 27
4.3 特徵大小實驗 ................ 30
4.4 RGB、Depth 及RGB-D 之比較 ... 32
4.5 合併及獨立之特徵比較.......... 33
4.6 PDF 及CDF 比對之比較 ........ 35
4.7 A3、D2 及D3 方法之比較 ...... 36
五、結論與未來展望 ............... 37
參考文獻 ........................ 38

[1]3DScanner,https://zh.wikipedia.org/wiki,2017.
[2]Microsoft,“Kinect for Windows,”http://www.kinectforwindows.org,2017.
[3]PCL‒Point Cloud Library (PCL),http://pointclouds.org/, 2017.
[4]Chi-Cheng Chen, “The Study of Image Matching Based on Feature Point Selection and Description Method,”National Kaohsiung University of Applied Sciences, 2013.
[5]Chen-Wen Lo,“Image Retrieval on Feature Point Selection,”International Conference on Advanced Information Technologies (AIT),2008.
[6]羅振文，植基於特徵點之影像檢索技術，玄奘大學資訊科學研究所碩士論文，2008年。
[7]黃育仁、李建平、陳冠良、李志忠、張美萍、黃秀菊，“Image Retrieval by Using Region Adjacency Relations,”東海科學, pp.17-41,民93.07
[8]Wan-HsinLi, “To Construct Block-based Bag of Visual Words,”I-Shou University, 2013.
[9]Ning Jiang, "Research on Gradient Local Binary Patterns Method for Human Detection," Waseda University Doctoral Dissertation, 2014.
[10]Bo-Rong Li, “LBP-Features Based License Plate Localization with Support Vector Machine,”National Pingtung University, 2017
[11]Yi-HongTseng and Chi-Wei Huang,“An Image Retrieval System Obtained by Integrating MultipleSearching Approaches,” Journal of Science and Engineering Technology, vol. 5, no. 3, pp. 1-12, 2009.
[12]R. Osada, T. Funkhouser, B. Chazelle and D. Dobkin, “Matching 3D Models with Shape Distributions,” Int. Conf. on Shape Modeling and Applications (SMI 2001), pp. 154-166, 2001.
[13]Chen-Kuo Lai,“A Content-Based Approach Towards 3D Model Retrieval,”Department of Information Engineering,I-Shou University,2004.
[14]Xiong Hua, “3D Model Retrieval Technique Based on Area Distribution,”Computer Engineering,vol. 5, no. 3,pp. 257-261, 2008.
[15]石昭玲、張浩澤, “3D Model Retrieval based on Local Discrete Fourier Transform,”Journal of Information Technology and Applications,vol. 9, no. 1, pp. 18-23, 2015.
[16]盧良鋒、何加銘、謝志軍等，基於深度學習的RGB-D物體識別算法，移動通信，vol. 39, no. 10, pp. 52-56, 2015.
[17]Wu Xin, Wang Guiying andCong Yang,“Object Recognition Method by Combining Color and Depth Information,” Transactions of the Chinese Society of Agricultural Engineering, vol. 29, pp. 96-100, 2013.
[18]牟家昌，結合二維影像與三維點雲資訊之物體辨識，國立交通大學電控工程研究所碩士論文，2011。
[19]Hang Su, Subhransu Maji, Evangelos Kalogerakis and Erik Learned-Miller, “Multi-view Convolutional Neural Networks for 3D Shape Recognition,”The IEEE International Conference on Computer Vision (ICCV), pp. 945-953, 2015.
[20]Kevin Lai, Liefeng Bo, Xiaofeng Ren and Dieter Fox,“A Large-Scale Hierarchical Multi-View RGB-D Object Dataset,”IEEE International Conference on Robotics and Automation (ICRA), May 2011.
[21]Visualization and Interactive Media Laboratory of NCHC, 透過OpenNI建立Kinect 3D Point Cloud, http://viml.nchc.org.tw/home/, 2007-2017.
[22]Liu Junjie, Qiao Wenbao, Shan Weibo and Yao Yuan,“Object Segmentation and Real-Time Reconstruction Approach Based on RGB-D Data,”Computer Applications and Software, vol. 32, no. 4, pp.215-221, 2015.
[23]Liangqiong Qu, Shengfeng He, Jiawei Zhang, Jiandong Tian, Yandong Tang and Qingxiong Yang, “RGBD Salient Object Detection via Deep Fusion,”IEEE Transactionson Image Processing, vol. 26, no. 5, May 2017.
[24]KouroshKhoshelham and Sander Oude Elberink,“Accuracy and Resolution of Kinect Depth Data for Indoor Mapping Applications,”Sensors, vol. 12, no. 2, pp. 1437-1454, 2012.
[25]Zhen-Ya Yang,YongWang andZhen-DongYang,“Vector-Angular Distance Color Difference Formula in RGB Color Space,”Computer Engineering and Applications, vol. 46, no. 6, pp. 154-156, 2010.
[26]Colour Metric, http://www.compuphase.com/cmetric.htm, 2017.