視障人士在生活諸多方面均面臨許多不便。只仰賴導盲手杖，穿越未設有導盲磚的馬路對於視障人士仍是一大挑戰。有鑑於此，本論文設計一套可攜式引導系統，藉由機器視覺技術，協助視障人士行走在行人穿越道的中央位置。因此，視障人士的安全性與自主性均可獲得改善。
本研究透過影像處理方法，縱使在行人遮蔽與陰影干擾的情況之下，仍能找到行人穿越道的中央位置。再搭配統計策略去除錯誤偵測的影響，並結合震動裝置，提供視障人士的引導資訊。實驗測試分成靜態與動態影像兩種，以檢視中線偵測之演算法與震動導引策略之結果成效。而為了考量行人穿越道在影像中所有的呈現狀況，將行人穿越道歸類成四種狀況：正對無遮蔽、斜對無遮蔽、正對有遮蔽、以及斜對有遮蔽。最後，所提出的輔助引導系統透過健康人遮蔽雙眼的穿越道路實際驗證，在引導其行走在行人穿越道上展現出具前景的成效

Visually impaired people always experience inconvenience to manage many issues in their daily life. Crossing the road without tactile paving becomes a great challenge to the blind, who can only rely on the tactile stick. Based on this motivation, the thesis aims to develop a portable assistive guiding system using techniques of machine vision to allow the visually impaired people can always walk on the central area of the crosswalks while crossing the road. Both safety and autonomy of visually impaired people can therefore be improved.
This research incorporates image processing approaches to locate the central position of the crosswalks even affected by occlusion of pedestrians and interference of shadow. In addition, statistical approaches are also applied to reduce the influence of fault detection. A vibration device is then employed to provide information of guidance for the visually impaired people. The experiments are divided into two groups, static and dynamic images, to examine performance of the proposed middle-line detection algorithm and guidance of vibration strategy. In order to consider all possible appearances of crosswalks in the image, four different conditions, head-on without occlusion, oblique without occlusion, head-on with occlusion, and oblique with occlusion, are studied. Finally, the presented assistive guiding system demonstrates promising performance to direct a healthy person with both eyes covered walking on the crosswalks to pass through a road in the real world.

目錄
論文審定書 i
誌謝 ii
摘要 iii
Abstract iv
目錄 v
圖次 viii
表次 xiii
第一章 緒論 1
1.1研究動機與目的 1
1.2研究方法與步驟 3
1.3文獻回顧 4
1.4論文架構 9
第二章 枕木紋行人穿越道之靜態影像 10
2.1測試影像蒐集 11
2.2色彩空間 12
2.2.1 RGB色彩空間 12
2.2.2 HSV色彩空間 13
2.3陰影偵測與修補處理 14
2.4白色特徵提取 16
2.5資訊保留與雜訊處理 19
2.5.1路面感興趣區域 19
2.5.2 面積篩選（Size Filter） 20
2.6枕木紋條紋線段偵測 23
2.6.1邊緣偵測 24
2.6.2霍夫轉換 27
2.7線段延伸與合併 30
2.7.1線段延伸 31
2.7.2線段合併 36
2.8斜率趨勢篩選 39
2.9數據統計之線段篩選 40
2.9.1四分位數篩選水平線段 41
2.9.2消失點篩選傾斜線段 42
2.10 找尋中央位置 44
2.10.1迴歸分析與線性模型 45
2.10.2最小平方法估計 45
2.10.3估計線性回歸之參數 47
第三章 枕木紋行人穿越道之動態影像 49
3.1中線結果策略 50
3.1.1 路面感興趣區域的大小檢測 51
3.1.2中線結果之統計 51
3.1.3錯誤結果的篩選 52
3.2震動策略 54
第四章 實驗結果 59
4.1硬體設備 59
4.1.1攝影機 60
4.1.2震動馬達 61
4.1.3 Arduino（UNO） 62
4.1.4系統整合 63
4.2靜態影像之測試結果 65
4.2.1場景設定 66
4.2.2測試影像 68
4.2.3數據分析 71
4.3動態影像之測試結果 83
4.3.1場景設定 84
4.3.2實驗步驟 86
4.3.3實驗結果 91
4.4真實場景之測試結果 100
第五章 結論與未來展望 102
5.1結論 102
5.2未來展望 103
參考文獻 105















圖次
圖1.1 行人與其他物體造成的遮蔽 3
圖1.2 光線照射物體產生陰影遮蔽..………………………………………………..3
圖1.3 系統架構圖 3
圖1.4 汽車結合機器視覺 4
圖1.5 道路偵測示意圖 4
圖1.6 號誌擷取示意圖 5
圖1.7 道路障礙示意圖 5
圖1.8 枕木紋行人穿越道偵測示意圖 6
圖1.9 可攜式引導系統示意圖 7
圖1.10 陰影對於綠地的遮蔽 8
圖1.11陰影對於交通標線的遮蔽……………………………………………………8
圖2.1 靜態影像處理流程圖 10
圖2.2 輸入之枕木紋行人穿越道影像 11
圖2.3進行處理之影像 11
圖2.4 RGB色彩空間表示圖 12
圖2.5 HSV色彩空間表示圖 13
圖2.6 影像轉換置HSV色彩空間 14
圖2.7 斑馬線上的陰影偵測與補償 15
圖2.8 路面偵測結果 16
圖2.9 白色特徵提取 17
圖2.10 白色提取結果 18
圖2.11 路面之感興趣區域 19
圖2.12 感興趣資訊保留 20
圖2.13 區塊聯通之遮罩 21
圖2.14 二值化影像示意圖 21
圖2.15 標籤化處理結果示意圖 22
圖2.16 面積篩選處理後結果 23
圖2.17 掃描之像素區塊 24
圖2.18 水平梯度遮罩..……………………………………………………………..24
圖2.19垂直梯度遮罩..…………………………………………………………..…..24
圖2.20 梯度方向之設定 26
圖2.21 滯後閾值篩選示意圖 27
圖2.22 霍夫平面轉換 28
圖2.23 霍夫平面轉換 29
圖2.24 霍夫轉換偵測線段結果 29
圖2.25 陰影與遮蔽導致枕木紋行人穿越道線段分段示意圖 30
圖2.26 圖像邊緣與偵測線段示意圖 31
圖2.27 影像空間示意圖 32
圖2.28 左端點沿斜率延伸 32
圖2.29 延伸點四捨五入 33
圖2.30 延伸點沿直線之垂直方向 33
圖2.31 邊緣檢查之三個參考點 34
圖2.32 延伸後的線段 34
圖2.33 左端點延伸之結果 35
圖2.34 線段延伸結果 35
圖2.35 期望合併之線段 36
圖2.36 偵測線段所屬類別 37
圖2.37 不合併線段之狀況 37
圖2.38 期望合併線段之狀況 38
圖2.39 線段合併之結果 38
圖2.40 斜率篩選 39
圖2.41 斜率趨勢篩選 40
圖2.42 四分位數表示圖 41
圖2.43 透視投影之消失點示意圖 42
圖2.44 消失點篩選偵測線段 43
圖2.45強影響值檢測並找尋消失點 44
圖2.46 迴歸曲線示意圖 46
圖2.47 殘差計算示意圖 46
圖2.48 中線偵測之結果 48
圖3.1 完整系流程圖 49
圖3.2 中線偵測之錯誤狀況 50
圖3.3 枕木紋行人穿越道末端之偵測結果不準確 51
圖3.4 當下影像錯誤的偵測結果 51
圖3.5 中線統計策略 52
圖3.6 新結果與舊結果比對 53
圖3.7 中線偵測結果之篩選與統計 53
圖3.8 統計中線（綠）與當下中線（紅）比較 54
圖3.9 使用者當下的方向示意圖 54
圖3.10 偵測中線之角度判斷 55
圖3.11 使用者方向與偵測中線平移差異之狀況 56
圖3.12 中心點與偵測中線之相對位置 56
圖3.13 系統作動時間關係圖 57
圖3.14 馬達震動策略流程圖………………………………………………………58
圖4.1 系統之硬體設備 59
圖4.2 Logitech HD C310之網路攝影機 60
圖4.3 震動馬達 61
圖4.4 震動馬達模組之支援平台 62
圖4.5 Arduino（UNO） 62
圖4.6 本論文之攜帶式系統 64
圖4.7 遮蔽視線用墨鏡 64
圖4.8 攝影機之架設 65
圖4.9 震動導引之系統 65
圖4.10 枕木紋行人穿越道示意圖 66
圖4.11 枕木紋行人穿越道之不同規格影像 67
圖4.12 枕木紋行人穿越道之不同狀況影像 68
圖4.13 原始影像進行中線偵測之結果（場景一） 69
圖4.14 原始影像進行中線偵測之結果（場景二） 70
圖4.15 行人穿越道之端點標示 71
圖4.16 行人穿越道之中線尋找 72
圖4.17 角度差與距離差之示意圖 73
圖4.18 兩直線之截距差異示意圖 73
圖4.19 偵測結果之數據分佈圖 75
圖4.20 依狀況劃分之數據分佈圖 76
圖4.21 數據分佈圖之部分放大圖 76
圖4.22 橫跨道路短 77
圖4.23 枕木紋條紋毀損 77
圖4.24 行人穿越道條紋超出影像範圍 78
圖4.25 與枕木紋條紋相似資訊太多 78
圖4.26 交通號誌燈遮蔽狀況之處理 81
圖4.27 建築物遮蔽狀況之處理 82
圖4.28 自製枕木紋行人穿越道之規格示意圖 83
圖4.29 模擬場景 84
圖4.30 枕木紋行人穿越道狀態之模擬場景 84
圖4.31 測試的起始位置與方向 85
圖4.32 受試者之起始視角 86
圖4.33 模擬各種狀況之起始位置 88
圖4.34 足跡紀錄圖 88
圖4.35 動態測試之終點 89
圖4.36 左、右腳足跡之記錄圖 90
圖4.37 行走軌跡之記錄圖 90
圖4.38 測試實驗之行走軌跡 92
圖4.39 正對且無遮蔽之行走軌跡結果 93
圖4.40 斜對且無遮蔽之行走軌跡結果 94
圖4.41 正對且有遮蔽之行走軌跡結果 95
圖4.42 斜對且有遮蔽之行走軌跡結果 96
圖4.43 震動與步伐對應圖 98
圖4.44 導引策略與步伐對應圖 99
圖4.45 模擬實驗之設置終點 100
圖4.46 正對有遮蔽之真實道路測試 101
圖4.47 斜對有遮蔽之真實道路測試 101








表次
表1.1 衛生福利部統計處對於視覺障礙者之統計 1
表4.1 Logitech HD C310攝影機之規格 60
表4.2 震動馬達之作動規格 61
表4.3 Arduino（UNO）之規格 63
表4.4 每個場景影像之偵測結果 74
表4.5 依行場景分類之統計數據 79
表4.6 依行人穿越道狀況分類之統計數據 80

[1] 內政部衛生福利部統計處：https://dep.mohw.gov.tw/DOS/lp-2976-113.html
[2] X. Luo, Y. Li, X. T. Ren, and J. J. Wang “Automatic road surface profiling with sensors fusion,” 2012 12th International Conference on Control, Automation, Robotics & Vision, Guangzhou, China, December 5-7, 2012, pp. 608-613.
[3] P. Foucher, Y. Sebsadji, J. P. Tarel, P. Charbonnier, and P. Nicolle, “Detection and recognition of urban road markings using images,” 2011 14th International IEEE Conference on Intelligent Transportation Systems, Washington, DC, USA. October 5-7, 2011, pp. 1747-1752.
[4] J. Choi, B. T. Ahn, and I. S. Kweon, “Crosswalk and traffic light detection via integral framework,” The 19th Korea-Japan Joint Workshop on Frontiers of Computer Vision, Incheon, South Korea, January 30- February 1, 2013, pp. 309-312.
[5] D. C. Hern´andez, A. Filonenko, D. Seo, and K. H. Jo, “Crosswalk detection based on laser scanning from moving vehicle,” 2015 IEEE 13th International Conference on Industrial Informatics, Cambridge, UK, July 22-24, 2015, pp. 1515-1519.
[6] A. Haselhoff and A. Kummert, “On visual crosswalk detection for driver assistance systems,” 2010 IEEE Intelligent Vehicles Symposium, University of California, San Diego, CA, USA, June 21-24, 2010, pp. 883-888.
[7] Y. Sebsadji, J. P. Tarel, P. Foucher, and P. Charbonnier, “Robust road marking extraction in urban environments using stereo images,” 2010 IEEE Intelligent Vehicles Symposium, University of California, San Diego, CA, USA, June 21-24, 2010, pp. 394-400.
[8] Y. Zhai, G. Cui, Q. Gu, L. Kong, “Crosswalk Detection Based on MSER and ERANSAC,” 2015 IEEE 18th International Conference on Intelligent Transportation Systems, Las Palmas, Spain, September 15-18, 2015, pp. 2770-2775.
[9] J. M. Loomis, R. G. Golledge and R. L. Klatzky, “Navigation system for the blind: Auditory display modes and guidance,” Presence, Vol. 7, No. 2, 1998, pp. 193-203.
[10] C. H. Lin, P. H. Cheng and, S. T. Shen, “Real-time dangling objects sensing: A preliminary design of mobile headset ancillary device for visual impaired,”, 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, USA, August 26-30, 2014, pp. 5723-2726.
[11] T. V. Matar´o, F. Masulli, S. Rovetta, A. Cabri, C. Traverso, E. Capris, and S. Torretta, “An assistive mobile system supporting blind and visual impaired people when are outdoor,” 2017 IEEE 3rd International Forum on Research and Technologies for Society and Industry, Modena, Italy, September 11-13, 2017.
[12] R. Tapu, B. Mocanu, and E. Tapu, “A survey on wearable devices used to assist the visual impaired user navigation in outdoor environments” 2014 11th International Symposium on Electronics and Telecommunications, Timisoara, Romania, November 14-15, 2014.
[13] K. Sato, A. Yamashita, K. Matsubayashi, “Development of a navigation system for the visually impaired and the substantiative experiment,” 2016 Fifth ICT International Student Project Conference, Nakhon Pathom, Thailand, May 27-28, 2016.
[14] T. Gonnot and J. Saniie, “Integrated machine vision and communication system for blind navigation and guidance,” 2016 IEEE International Conference on Electro Information Technology （EIT）, May 19-21, 2016, pp. 187-191.
[15] A. R. García, R. Fonseca, and A. Durán “Electronic long cane for locomotion improving on visual impaired people. A case study,” 2011 Pan American Health Care Exchanges, Rio de Janeiro, Brazil, March 28-April 1, 2011, pp. 58-61.
[16] 林宗翰，利用 RFID 感測器於室內環境之盲人導航系統，國立臺灣師範大學機電科技學系碩士論文，2012年7月。
[17] M. C. Ghilardi, J. C. S. Jacques Jr., and I. H. Manssour, “Crosswalk localization from low resolution satellite images to assist visually impaired people,” IEEE Computer Graphics and Applications （ Volume: PP, Issue: 99 ）, May 25, 2016.
[18] V. N. Murali and J. M. Coughlan, “Smartphone-based crosswalk detection and localization for visually impaired pedestrians,” 2013 IEEE International Conference on Multimedia and Expo Workshops, San Jose, CA, USA, July 15-19, 2013.
[19] M. Radványi, B. Varga, and K. Karacs “Advanced crosswalk detection for the bionic eyeglass,” 2010 12th International Workshop on Cellular Nanoscale Networks and Their Applications, Berkeley, CA, USA, February 3-5, 2010.
[20] 范賀翔，枕木紋行人穿越道偵測之影像處理技術，國立中山大學機械與機電工程學系碩士論文，2010年6月。
[21] 梁晉瑋，應用機器視覺技術於行人穿越道線與行人專用號誌辨識之研究，國立台北科技大學碩士學位論文，2010年7月。
[22] T. Asami and K. Ohnishi, “Crosswalk location, direction and pedestrian signal state extraction system for assisting the expedition of person with impaired vision,” 10th France-Japan Congress, 8th Europe-
Asia Congress on Mecatronics, Tokyo, Japan, November 27-30, 2014, pp. 285-290.
[23] C. Blajovici, P. J. Kiss, Z. Bonus, and L. Varga, “Shadow detection and removal from a single image,” 19th Summer School on Image Processing, Szeged, Hungary, July 7-16, 2011.
[24] L. Powell and K. G. Satheeshkumar, “Automated road distress detection,” 2016 International Conference on Emerging Technological Trends, Kollam, India, October 21-22, 2016.
[25] V. Jain and A. Khunteta, “Shadow removal for umbrageous information recovery in aerial images,” 2017 International Conference on Computer, Communications and Electronics, Jaipur, India, July 01-02, 2017, pp. 536-540.
[26] T. Shioyama, H. Wu, Y. Nishibe, N. Nakamura, and S. Kitawaki, “Image analysis of crosswalk,” 11th International Conference on Image Analysis and Processing, September 26-28, 2001, Palermo, Italy, pp. 168-173.
[27] N. Hayashi, T. Tomizawa, T. Suehiro, S. Kudoh, “A robust white line detection technique for double circular operator,” 2012 International Conference on Mechatronics and Automation, Chengdu, China, August 5-8, 2012, pp. 1949-1954.
[28] S. Wang and Y. Tian, “Detecting stairs and pedestrian crosswalks for the blind by RGBD camera,” 2012 IEEE International Conference on Bioinformatics and Biomedicine Workshops, Philadelphia, PA, USA, October 4-7, 2012, pp. 732-739.
[29] 黃奕和，基於平行線系的隨機霍夫轉換直線偵測方法，國立台北科技大學資訊工程系碩士班碩士學位論文，2016年1月。
[30] T. Zhang, H. M. Hu and B. Li, “A naturalness preserved fast dehazing algorithm using HSV color space,” IEEE Access（ Volume: 6 ）, February 15, 2018.
[31] J. Canny, “A computational approach to edge detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, （ Volume: PAMI-8, Issue: 6,）pp. 679 – 698, Nov, 1986.
[32] P. V. C. Hough, “Method and means for recognizing complex patterns,” U.S. Patent 3,069,654, December 18, 1962
[33] R. D. Cook, “Detection of influential observation in linear regression,” Technometrics, American Statistical Association and American Society for Quality, vol. 19, No. 1, Feberary, 1977, pp. 15-18
[34] J. S. Milton and J. C. Arnold, “Introduction to probability and statistics: principles and applications for engineering and the computing sciences,” McGraw-Hill Education, 4th edition, September 30, 2002.
[35] http://law.moj.gov.tw/LawClass/LawAll.aspx?PCode=K0040014