系統識別號 | U0002-0507201114470000 |
---|---|
DOI | 10.6846/TKU.2011.00154 |
論文名稱(中文) | 藉由弧探索改善泡泡探索 |
論文名稱(英文) | Bubble detection improved by using arc detection |
第三語言論文名稱 | |
校院名稱 | 淡江大學 |
系所名稱(中文) | 資訊工程學系碩士班 |
系所名稱(英文) | Department of Computer Science and Information Engineering |
外國學位學校名稱 | |
外國學位學院名稱 | |
外國學位研究所名稱 | |
學年度 | 99 |
學期 | 2 |
出版年 | 100 |
研究生(中文) | 安久知弘 |
研究生(英文) | Tomohiro Ankyu |
學號 | 698411260 |
學位類別 | 碩士 |
語言別 | 英文 |
第二語言別 | |
口試日期 | 2011-06-19 |
論文頁數 | 63頁 |
口試委員 |
指導教授
-
洪文斌(horng@mail.tku.edu.tw)
委員 - 謝文恭(wgshieh@faculty.pccu.edu.tw) 委員 - 范俊海(chunhai@mail.tku.edu.tw) 委員 - 洪文斌(horng@mail.tku.edu.tw) |
關鍵字(中) |
弧探索 圓探索 邊緣探索 |
關鍵字(英) |
Arc Detection Circle Detection Edge Detection |
第三語言關鍵字 | |
學科別分類 | |
中文摘要 |
在這份論文裡,各種大小的弧形都將被我們所提出的方法正確偵測出來。這是此份研究最重要的部分。將全部的弧形偵測出來後,我們還要估算出圓的總面積,因為這份研究的主要目的在於估算出泡泡在風力發電機表面上的總面積。 即使這項研究的主要目的是要估算出所有在風力發電機上的泡泡的總面積,在這份論文中,估算完泡泡總面積後,我們還會加入圓形偵測的程序來描繪出泡泡的輪廓。 我們增加程序來繪出泡泡輪廓是因為過程中有發生測量的錯誤,即使測量的錯誤接受度很低。事實上,我們能夠加以運用描繪泡泡輪廓的過程計算出正確的面積。我們需要運用圓周偵測來偵測圓形的輪廓。 我們提到的方法圓形偵測的是由Hough transform來處理。我們在堆疊中找到相對的高映像點做為圓心候補。然後我們再畫個圓心候補的的圓形。 |
英文摘要 |
In this paper, whatever sizes of circles are small or large, all arcs of circles on images are able to be detected by our proposed method, correctly. This is the most important point of this research. After detecting arcs, we are going to evaluate total area of circles on surface of image, because the main purpose of this research is to attribute to evaluate total area of bubbles on surface of wind turbine. Even though the main purpose of this research is to evaluate total area of all bubbles on surface of wind turbine, in this thesis we will add the processes to draw contour of bubble by circle detection after evaluating total area of bubbles. The reason which we add the processes to draw contour is the processes have measurement error even though the measurement error is acceptable low. In fact, we can calculate accurate area by applying the process of drawing contour of bubble. We need to apply circle detection to detect contour of circle. Our proposed method of circle detection is by Hough transform. We detect the comparatively high pixels of accumulation as center candidates. Then we draw circles after finding the center candidates. In addition to evaluating accurate area, the proposed method has advantages like the followings. By our original method, it is also possible to detect circles which have shortages, overlapped parts, theoretically. It is difficult to detect edges and area and contours of circles which have different sizes. We demonstrate accuracy of our proposed method by comparing our proposed methods with other possible methods. |
第三語言摘要 | |
論文目次 |
Contents Chapter 1 Introduction 1 Chapter 2 Related Work 4 2.1 Circle Detection 4 2.1.1. Real-time Robust Algorithm for Circle Object Detection 4 2.2 Corner Detection 6 2.2.1. Robust Detection of Corners and Corner-line Links in Images 6 2.2.2. Robust Image Corner Detection Using Local Line Detector 7 Chapter 3 Terminology 9 3.1 Sobel Filter 9 3.2 Hough Transform 11 3.3 Hough-based schemes for Circular object Detection 14 3.4 Inverse 15 3.5 Canny Filter 18 3.6 Otsu's Method 21 3.7 Contrast Stretching 23 3.8 Binarization 25 Chapter 4 Actual Implementation 26 4.1 Our Proposed Method 26 4.1.1. Edge Detection 26 4.1.2. Circle Detection 31 4.2 Other Methods 32 4.2.1. Method1 32 4.2.2. Method2 34 4.2.3. Method3 35 4.2.4. Method4 37 Chapter 5 Experiments 45 5.1 Our Proposed Method( Edge Detection ) 45 5.1.1. Result of Edge Detection 45 5.1.2. Result of area 46 5.2 Our Proposed Method ( Circle Detection ) 47 Chapter 6 Conclusion 52 Future Work 54 References 55 Appendix : English Paper 56 List of Figures Figure 2.1 The validation of an arc by splitting the art into two sub-arcs across the straight line passing through both arc’s gravity center (xm,ym) and circular center (xc,yc) 5 Figure 2.2 Steps of corner detection 6 Figure 3.1 Original image(left), the result of Sobel filter(rihgt) 11 Figure 3.2 The definitions of ρ,θ 12 Figure 3.3 Combinations of distance and angle 13 Figure 3.4 The binary image of arcs 15 Figure 3.5 The result of Hough transform 16 Figure 3.6 The result of drawing circles by their center candidates 17 Figure 3.7 Compare original circles(Black) with the results of Hough transform 17 Figure 3.8 The result of Hough transform 18 Figure 3.9 The Gaussian mask 3×3 19 Figure 3.10 The Gaussian mask 5×5 19 Figure 3.11 Original image(left), the result of Canny filter(right) 21 Figure 3.12 Otsu method 21 Figure 3.13 Original image(left), the result of Otsu(right) 23 Figure 3.14 Original image(left), the result of Contrast stretching(right) 24 Figure 3.15 Original image(left), the result of Binarized image(right) 25 Figure 4.1 Original images 27 Figure 4.2 The results of Edge emphasize and Contrast emphasize 27 Figure 4.3 The result of luminance(grey pixel as a bench mark) 28 Figure 4.4 The result of first derivative of Figure 4.17 29 Figure 4.5 Compare Figure 4.17(luminance) with Figure 4.18(first derivative) 29 Figure 4.6 The result of second derivative of Figure 4.18 29 Figure 4.7 Compare Figure 4.18(first derivative) with Figure 4.20(second derivative) 30 Figure 4.8 Compare Figure 4.18(first derivative) ,Figure 4.20(second derivative) with Figure 4.17(luminance) 30 Figure 4.9 The result of Contrast Stretching + Sobel filter 33 Figure 4.10 The result of Contrast Stretching + Sobel filter + Binarization 33 Figure 4.11 The result of Canny detection 35 Figure 4.12 The result of Sobel filter + substracting background + Canny detection 36 Figure 4.13 The result of Otsu 36 Figure 4.14 Pixels of edge p1, p2 37 Figure 4.15 Pixels of edge, p1, p2, p3 38 Figure 4.16 Center candidate detection by lining perpendiculars of p1,p3 and p2,p3 38 Figure 4.17 The result of applying our proposed method to overlapped circles 40 Figure 4.18 The result of drawing circles by the method of Figure 4.24 40 Figure 4.19 The result of repeating 32 times 41 Figure 4.20 The result of repeating 64 times 42 Figure 4.21 The result of repeating 96 times 42 Figure 4.22 The ideal result of our proposed method 43 Figure 4.23 The result of applying our proposed method to the image which includes overlapped bubbles 43 Figure 4.24 The ideal result of applying our proposed method to overlapped bubbles 44 Figure 5.1 The result of our proposed arc detection 45 Figure 5.2 The result of applying our proposed method to overlapped bubbles 46 Figure 5.3 The result of painting area of bubbles 46 Figure 5.4 The result of painting area overlapped bubbles 47 Figure 5.5 The result of Binary image of Our proposed arc detection 47 Figure 5.6 The result of Hough transform setting radius 8-10 48 Figure 5.7 The result of Hough transform setting radius 20-25 49 Figure 5.8 The result of Hough transform setting radius 30-35 49 Figure 5.9 The result of Hough transform and project setting radius9-10 50 Figure 5.10 The result of Hough transform and project setting radius11-12 50 Figure 5.11 The result of Hough transform and project setting radius6-7 51 Figure 5.12 The result of Hough transform and project setting radius8-9 51 |
參考文獻 |
[1] Weili Ding, Xiaoli Li, and Wenfeng Wang, "Robust Image Corner Detection Using Local Line Detector and Phase Congruncy Model," in WASE International Conference on Information Engineering, 2010, pp. 158-161. [2] Jianping Wu, Jinxiang Li, Changshui Xiao, Fangyong Tan, and Caidong Gu, "Real-time Robust Algorithm for Circle Object Detection," in The 9th International Conference for Young Computer Scientists, 2008, pp. 1722-1727. [3] Andres Solis Montero, Milos Stojmenović, and Amiya Nayak, "Robust Detection of Corners and Corner-line links in images," in 10th IEEE International Conference on Computer and Information Technology (CIT 2010), 2010, pp. 495-502. |
論文全文使用權限 |
如有問題,歡迎洽詢!
圖書館數位資訊組 (02)2621-5656 轉 2487 或 來信