臺灣博碩士論文加值系統

盲人因為視力損傷的緣故，對環境的探索遠比健全人士吃力，在生活上往往會遇到許多不便和危險，因此本論文介紹一個基於Kinect影像的室內盲人防碰撞系統，希望透過此系統幫助視障者在室內環境中，可以避開有可能撞上的障礙物。
本系統透過讀取Kinect的深度影像和彩色影像，並且針對深度影像進行影像處理、物件偵測、障礙物辨識，最後將障礙物的距離和方位透過螢幕與振動器提供給使用者。以影像金字塔的方法來進行影像降取樣以提升系統執行效率，並以中值濾波器平滑深度影像。透過區域成長來進行物件偵測以得到候選障礙物，接著以四個原則來過濾候選障礙物，以得到障礙物及其資訊。同時，也在論文中探討區域成長使用在深度影像中的種子數量和擺設位置，以獲得畫面中所有的物件，並分別在簡單場景與複雜場景進行測試，最後以20個種子數量平均擺設在深度影像中有最佳搜尋候選障礙物的結果。
總合來說，本研究提出一套基於Kinect影像的高效能室內盲人防碰撞輔助系統，以每0.3秒左右處理一張影像，並能將所得障礙物方位和距離資訊透過振動器以不同的頻率來提示使用者。
關鍵字: 盲人輔助系統、Kinect Sensor、區域成長

The visually impaired people usually need to struggle in their daily lives to explore their outside environment. If more information of the environment can be provided in some way with the assistance of recent technology, then they could adapt to the environment more rapidly. This paper presents an anti-collision blind aid system in the indoor environment based on the Kinect sensor. The system aims to assist the visually impaired avoiding form the obstacles in the indoor environments.
The system obtains the environment information by Kinect sensor, which generates the color image and the depth image. With the image downsampling, image processing, object detection, and obstacle detection of the depth image, the user is able to feel the distance and direction of the obstacles by the vibration module. To promote system efficiency, we downsample the image by image pyramid method. After that, we use median filter to eliminate noise and smooth the image. We obtain the obstacle candidate by using region growing in depth image and using four principles to distinguish the obstacles or not. In the thesis, we also study the issue of the number and positions of the seeds in the region growing method in order to obtain all the objects in the depth image. We tested in both simple and complex scenes and determined that using 20 seeds would achieve the best result.
As the result, we developed an efficient indoor anti-collision blind aid system based on Kinect processing speed of obstacle detecting is 0.3 second per frame. By using this system, the visually impaired is able to know the position and the distance of the obstacle through the different vibration frequency form the vibration module.
Key Words:Blind Aid System, Kinect Sensor, Region Growing

致謝 I
摘要 II
ABSTRACT III
目錄 IV
圖目錄 VII
表目錄 X
第一章 緒論 1
1.1研究背景與動機 1
1.2研究目的 1
1.3論文架構 2
第二章 相關文獻回顧 3
2.1障礙物偵測 3
2.2 Xbox360 Kinect 5
2.2.1 OpenNI SDK 2 6
2.3 Arduino 8
2.3.1 Arduino Duemilanove 9
2.3.2震動馬達模組 11
第三章 系統架構及方法 12
3.1 系統架構 12
3.2 影像擷取 13
3.3 影像降取樣 13
3.4 影像處理 14
3.4.1中值濾波器(Median filter) 15
3.4.2圖像修補(Inpainting) 16
3.5 物件偵測 20
3.5.1區域分割定義 20
3.5.2區域成長法 20
3.5.3區域成長演算法 23
3.6 障礙物偵測 26
3.7 顯示系統 28
3.8 震動輸出系統 29
第四章 研究結果與討論 30
4.1 Kinect距離測試 30
4.2 降取樣效果 32
4.3 種子點的擺設和個數 33
4.3.1 簡單場景 33
4.3.2 複雜場景 35
4.4偵測障礙物與室內環境測試 38
4.4.1走廊與狹窄走道 38
4.4.2牆與密閉環境 40
4.4.3明亮環境與黑暗環境 41
4.5討論 44
第五章 結論 46
第六章 未來展望 47
第七章 參考文獻 48

[1] World Health Organization Website, Visual impairment and blindness, 2012, http://www.who.int/mediacentre/factsheets/fs282/en/
[2] 侯志平 ,“應用單一攝影機影像於居家及醫療照護場所之跌倒監測”, 長庚大學, 2010
[3] S. Meers and K.Ward, “Face recognition using a time-of-light camera," in Computer Graphics, Imaging and Visualization, 2009. CGIV '09. Sixth International Conference on, pp. 377 -382, aug. 2009.
[4] Wannes van der Mark and Dariu M. Gavrila, “Real-Time Dense Stereo for Intelligent Vehicles”, IEEE Transactions on Intelligent Transportation Systems, VOL. 7, NO. 1, MARCH 2006
[5] Don Murray and Cullen Jennings, “Stereo vision based mapping and navigation for mobile robots”, Proceedings of the 1997 IEEE International Conference on Robotics and Automation Albuquerque, New Mexico - April 1997
[6] D T Delpy, M Cope, P van der Zee, S Arridge, S Wray and J Wyatt , “Estimation of optical pathlength through tissue from direct time of flight measurement”, Phys. Med. Biol., Vol. 33, No 12, 1433-1442, 1988
[7] 李嘉祥, “針對盲人輔助應用之電腦視覺演算法開發與硬體架構設計”, 國立臺灣大學電機資訊學院電子工程學研究所碩士論文, 2012
[8] 蘇仕瑋, “以Kinect配合智慧型手機實做盲人輔具系統”, 國立成功大學工科科學學系碩士論文, 2013
[9] Steve Mann, Jason Huang, Ryan Janzen, Raymond Lo, Valmiki Rampersad, Alexander Chen, Taqveer Doha, “Blind navigation with a wearable range camera andvibrotactile helmet”, Scottsdale, Arizona, USA, MM’11 November 28–December 1, 2011
[10] Michael Zöllner, Stephan Huber, Hans-Christian Jetter, Harald Reiterer, “NAVI –AProof-of-Concept of a Mobile Navigational Aid for Visually Impaired Based on the MicrosoftKinect”, University of Konstanz, 2011.
[11] Jinwook Choi, Deukhyeon Kim, Hunjae Yoo, and Kwanghoon Sohn, “Rear Obstacle Detection System Based on Depth From Kinect”, 2012 15th International IEEE Conference on Intelligent Transportation Systems, pp. 98-101, September, 2012
[12] Kinect for window, http://www.microsoft.com/en-us/kinectforwindows/
[13] OpenNI, http://www.openni.org/
[14] OpenNI, http://www.openni.org/about/#.Ub4Cq-f7B8E
[15] Arduino簡介, http://www.robotsky.com/e/DoPrint/?classid=60&id=4346
[16] Arduino Duemilanove, http://arduino.cc/en/Main/arduinoBoardDuemilanove
[17] Edge detection with image pyramid, http://fourier.eng.hmc.edu/e161/lectures/canny/node2.html
[18] 紹謬綱, “數位影像處理”, 台灣培生教育出版股份有限公司, 2009
[19] Alexandru Telea ,“An Image Inpainting Technique Based onthe Fast Marching Method”, Journal of Graphics Tools 9(1), ACM, Vol. 9, No. 1: 25—36 2004
[20] A. Lejeune, S. Piérard, M. Van Droogenbroeck, J. Verly, “A new jump edge detection method for 3D cameras”, Conference on 3D Imaging (IC3D), Liège, Belgium, December 2011
[21] Wikipedia, Breadth-First-Search, http://zh.wikipedia.org/wiki/File:Breadth-first_tree.svg
[22] 透過 OpneNI 讀取 Kinect 深度影像資料, http://kheresy.wordpress.com/2011/01/20/read_kinect_depth_data_via_openni/