由於現今車載系統中，針對道路行車速限的提示，一般方法常使用 GPS 定位資訊搭配事前建立的資料庫，搜尋該路段行車速限紀錄，以提示駕駛注意行車的速限，但此方法往往因資料過舊或道路的臨時變更，而導致錯誤的提示。本文提出一個能偵測與辨識道路速限標誌的方法，並且實作出一套視覺行車輔助系統，其為結合行車記錄器所拍攝之影像，在車輛行進間，即時的辨識出道路速限標誌之道路速限。針對不同角度及距離所偵測到的道路速限標誌，為提高此行車輔助系統的分類辨識成功率，本文亦提出一種特徵描述的方法流程，其結合影像質心和輪廓距離角度的特徵計算方法，能在有限的角度範圍內，達到平移不變、尺度不變、旋轉不變的特徵描述方式，作為支持向量機的分類的數據資料。接著透過大量資料庫的建立，讓支持向量機分類的能力提高。最後，將訓練好的分類器移植至Radxa  Rock  Pro 嵌入式平台上，其配備 ARM Cortex-A9 1.6GHz 四核心的處理器，可運行 Android 4.4.2 的作業系統，並移植OpenCV 和 LibSVM 的函式庫，在 JNI 層實作本論文所提出的影像辨識系統，對嵌入式平台效能做程式碼的優化，以達到即時偵測與辨識出道路的速限之目的。

In a vehicle navigation system, the design of road-speed-limit warning function is often implemented by using global positioning system combined with a pre-established traffic sign database to search the recorded road-speed-limit sign close to the current location. However, this method may obtain a wrong advice result when the database does not update to the latest version. This thesis addresses the design and implementation of a vision-based  driver-assistance system based on an image-based road-speed-limit sign recognition algorithm, which is able to automatically detect and recognize the road-speed-limit sign on the road in real-time. To improve the recognition rate of road-speed-limit signs with different view angles and scales, this thesis also presents a new feature extraction algorithm, which describes an object using the distances between the object’s image centroid and its contour under some specific angle conditions. The proposed feature descriptor achieves translation, scale, and rotation invariant with respect to a certain view angle range. This property helps to improve the recognition accuracy of a support vector machine classifier trained by using a large database of traffic signs. The proposed vision-based driver-assistance system had been implemented on a Radxa Rock Pro embedded platform equipped with an ARM Cortex-A9 Quad Core 1.6GHz CPU running Android 4.4.2 operating system. The proposed system was implemented in JNI layer using OpenCV and LibSVM libraries, allowing achieving real-time performance by C code optimization.

目錄
中文摘要	I
Abstract	II
目錄	III
圖目錄	VI
表目錄	IX
第一章	 序論	1
1.1	 相關研究	1
1.2	 研究動機與目的	3
1.3	 論文架構	4
第二章  實驗系統介紹	5
2.1	 軟硬體平台介紹	5
2.2	 OpenCV	6
2.3	 LibSVM	7
第三章	 相關背景知識	8
3.1	 影像矩	8
3.2    HSV的色彩空間模型	10
3.3    Otsu自動閥值	11
3.4	 連通元件標記	12
第四章	 道路速限標誌偵測與辨識系統	14
4.1    標誌偵測處理	16
4.1.1  影像色彩空間轉換	17
4.1. 2  速限標誌偵測演算法	20
4.2    標誌ROI提取	20
4.2.1  Otsu閥值計算與二值化轉換	21
4.2.2  四連通方法ROI提取	22
4.3    標誌辨識處理	24
4.3.1  質心座標與影像方向角計算	24
4.3.2  輪廓特徵搜尋與計算	26
4.3.3  影像方向角補償	28
4.3.4  輪廓特徵正規化及直方圖統計	28
4.3.5  支持向量機分類預測	30
第五章  實驗結果與分析	31
5.1    影像方向角分析	31
5.2    特徵描述方法比較	33
5.3    系統實測結果	41
5.4    系統效能分析	49
第六章  結論與未來展望	51
參考文獻	52
附錄A		55
附錄B		57















 
圖目錄
圖3.1、HSV的色彩空間模型	10
圖3.2、連通方向示意圖 (a) 四連通的方法 (b) 八連通的方法	13
圖4.1、系統方塊圖	14
圖4.2、特徵抽取示意圖[22] (a) 36個方向標準圖特徵抽取示意圖 (b) 特徵值抽取順序圖	15
圖4.3、系統流程圖	16
圖4.4、白天影像擷圖(a) 影像原圖 (b) 僅以H<15° or H>330°且Ｓ＞0.35做為閥值條件 (c) 增加V＞0.35閥值條件	18
圖4.5、夜晚影像擷圖(a) 影像原圖 (b) 增加V＞0.35閥值條件 (c) 改以V＞0.15做為閥值條件	19
圖4.6、藍色矩形框選的範圍表示的後續要識別的資料 (a) 白天影像中有3處被選取 (b) 夜晚影像中有9處被選取	20
圖4.7、閥值選擇的影響 (a) 實際圖 (b) 灰階反轉圖 (c) Otsu自動閥值 (d) 閥值過大 (e) 閥值過小	21
圖4.8、可辨識的道路速限標誌 (a) 道路速限標誌 (b) 四連通分割後的ROI	22
圖4.9、其他類別的交通標誌	23
圖4.10、待排除的標誌ROI	23
圖4.11、不同的角度及距離拍攝到的速限標誌示意圖 (a) 標誌距離過近 (b) 標誌大小適中 (c) 標誌傾斜	24
圖4.12、ROI裁切後的結果示意圖 (a) 正常情況 (b)平移 (c)傾斜 (d)過小	25
圖4.13、P_1 (輪廓上之任一點)與質心P_c相對座標之關係圖 (a) P_1 (73,49),P_c (50.83,44.95) (b) P_1 (84,64),P_c (61.83,59.95)	25
圖4.14、四方向檢測示意圖 (a)檢查P_1是否為輪廓點 (b)搜尋結果	27
圖4.15、計算γ值與θ值	27
圖4.16、排列後的角度與距離直方圖	28
圖4.17、正規化ROI的示意圖	29
圖5.1、方向角差(a) 速限30標誌內容 (b) 速限40標誌內容 (c) 速限50標誌內容 (d) 速限60標誌內容 (e) 速限70標誌內容 (f) 速限80標誌內容 (g) 速限90標誌內容 (h) 速限100標誌內容(i) 速限110標誌內容	33
圖5.2、影像資料庫示意圖	34
圖5.3、F-度量統計各類別的辨識率 (a) 資料庫一資料集合(線性內核)  (b) 資料庫一資料集合(多項式內核)  (c) 資料庫一資料集合(高斯核)  (d) 資料庫二資料集合(線性內核)  (e) 資料庫二資料集合(多項式內核)  (f) 資料庫二資料集合(高斯核)	39
圖5.4、資料庫中同類別不同尺寸的圖片 (a) 95×96 (b) 69×68 (c) 46×46 (d) 30×30 (e) 23×20	40
圖5.5、辨識結果輸出示意圖 (a) 用來表示未知的類別 (b) 用來表示ROI過小	41
圖5.6、系統實測結果 (a) 白天實測結果 (b) 夜晚實測結果 (c) 偵測偏差但仍可辨識成功的案例 (d) 標誌歪斜但仍可辨識成功的案例 (e) 拍攝距離較遠但仍可辨識成功的案例 (f) 偵測失敗的案例 (g) 辨識失敗的案例	47
圖5.7、HSV轉換色彩閥值的結果	48
 
表目錄
表2.1、開發板規格表[14]	6
表2.2、攝影機模組規格表	6
表5.1、交叉驗證辨識結果	35
表5.2、用於評估資訊檢索系統的成效之列表[23]	36
表5.3、資料集合與預測結果分類統計 (a) 資料庫一資料集合分類統計(b) 資料庫二資料集合分類統計	37
表5.4、系統各階段處理時間	49

參考文獻
[1]	N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, USA, Vol. 1, pp. 886-893, 2005.
[2]	D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” International Journal of Computer Vision, Vol. 60, No. 2, pp. 91-110, 2004.
[3]	H. Bay, T. Tuytelaars, and L. V. Gool, “SURF：speeded up robust features,” Proceedings of 9th European Conference on Computer Vision, Graz, Austria, pp. 404-417, 2006.
[4]	J. MacQueen, “Some methods for classification and analysis of multivariate observations,” Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability 1. University of California Press, pp. 281-297, 1967.
[5]	T. M. Cover and P. E. Hart, “Nearest neighbor pattern classification,” IEEE Transactions on Information Theory, Vol. 13, No. 1, pp. 21-27, 1967.
[6]	C. Cortes and V. Vapnik, “Support-vector networks,” Machine Learning, Vol. 20, No. 3, pp.273-297, 1995.
[7]	A. Martinović, G. Glavaš, M. Juribašic, D. Sutic, and Z. Kalafatic, “Real-time detection and recognition of traffic signs,” Proceedings of 35th International Convention on Information and Communication Technology, Electronics and Microelectronics, Opatija, Croatia, pp. 760-765, 2010.
[8]	N. Bartneck and W. Ritter, “Color segmentation with polynomial classification,” Proceedings of 11th International Conference on Pattern Recognition Methodology and Systems, The Hague, Vol. 2, pp. 635-638, 1992.
[9]	K.A. Ishak, M.M. Sani, and N.M. Tahir, “A Speed limit sign recognition system using artificial neural network,” 4th Student Conference on Research and Development, Shah Alam, Selangor, Malaysia, pp. 127-131, 2006
[10]	Y. B. Damavandi and K. Mohammadi, “Speed limit traffic sign detection and recognition,” IEEE Conference on Cybernetics and Intelligent Systems, Singapore, Vol. 2, pp. 797-802, 12. 2004.
[11]	M. Á. Garcia-Garrido, M. Á. Sotelo, and E. Martín-Gorostiza, “Fast road sign detection using hough transform for assisted driving of road vehicles,” Proceedings of 10th International Conference on Computer Aided Systems Theory, Las Palmas de Gran Canaria, Spain, pp. 543-548, 2005.
[12]	J. Torresen, J.W. Bakke and L. Sekanina, "Efficient recognition of speed limit signs," Proceedings of 7th International IEEE Conference on Intelligent Transportation Systems, pp. 652-656, 2004.
[13]	W. Liu, Y. Liu, H. Yu, H. Yuan and H. Zhao, "Real-time speed limit sign detection and recognition from image sequences," International Conference on Artificial Intelligence and Computational Intelligence, Sanya, Vol. 1, pp. 262-267, 2010.
[14]	Radxa Rock-Pro platform hardware specification at: http://radxa.com/Rock/specifications.
[15]	C. C. Chang and C. J. Lin, “LIBSVM: a Library for support vector machines,” ACM Transactions on Intelligent Systems and Technology, Vol. 2, No. 3, article 27, 2011.
[16]	R.E. Fan, P.H. Chen and C.J. Lin, “Working set selection using second order information for training SVM,” Journal of Machine Learning Research, Vol. 6, pp. 1889-1918, 2005.
[17]	M. K. Hu, “Visual pattern recognition by moment invariants,” IEEE Transactions on Information Theory, Vol. 8, No. 2, pp. 179-187, 1962.
[18]	HSV的色彩空間模型圖示: http://en.wikipedia.org/wiki/HSL_and_HSV#/media/File:HSV_color_solid_cylinder_alpha_lowgamma.png
[19]	Y.-Q. Wang, L.-M. Liu and Y. Zhao, “Traffic sign detection based on fixed color combination and intensity restraint,” International Symposium on Computer Network and Multimedia Technology, Wuhan, pp. 1-5, 2009. 
[20]	F. Ren, J. Huang, R. Jiang, and R. Klette, “General traffic sign recognition by feature matching,” 24th International Conference on Image and Vision Computing New Zealand, Wellington, pp. 409 – 414, 2009.
[21]	N. Otsu, “A threshold selection method from gray-level histograms," IEEE Transactions on Systems, Man and Cybernetics, Vol 9, No. 1, pp. 62 – 66, 1979
[22]	Jen-Hung Tsao, “Real-time traffic sign recognition technology,” National digital library of theses and dissertations in Taiwan, 2013.
[23]	評估資訊檢索系統的成效表: http://terms.naer.edu.tw/detail/1679003/