全球定位系統 (GPS) 技術的應用相當的普遍，但衛星訊號易受大樓等建築物的干擾或地形的遮蔽，且無法在室內使用。本論文以不使用GPS的情況下，提出一利用視覺標誌（Visual Beacon）定位，達到差動式兩輪機器人（Differential Wheeled Robot）自主導航之目的。本論文使用加速強健特徵（Speeded Up Robust Features, SURF）辨識標誌，但為了加速辨識速率與提升辨識率，在辨識之前，依據視覺標誌的特徵做了前置的影像處理。
本研究在本校園內實驗，根據校園的環境特徵，訂出了各具特色的視覺標誌。會預先偵測視覺標誌其特徵點 (interest points) 和擷取其描述子 (descriptors)，並儲存至平板的記憶體中，以方便後續比對使用，並節省了重複對視覺標誌偵測其特徵點和擷取其描述子的時間。在辨識視覺標誌前，先以視覺標誌在影像中可能出現的區域，選取較大範圍的感興趣區域（Region of Interest, ROI）。若視覺標誌具有顯著的顏色特徵，再以顏色偵測出真正的感興趣區域。這樣不僅可加快影像處理速度，並增加了辨識率。另外，由於校園中樹木林立、樹葉茂盛，存在了大量的特徵點，不僅增加影像處理時間，且這些特徵點並非固定，所以無法使用。因此本論文之視覺系統透過顏色偵測去除樹葉，不僅節省了大量運算時間在偵測特徵點和擷取描述子，亦提升了辨識率。

Global Positioning System (GPS) technology is fairly universal, but the satellite signal sometimes is not reliable by interference from the tall buildings, and can not be used indoors. In this thesis, we propose to use the visual beacons to position, instead of using GPS. The proposed method can achieve a differential wheeled robot autonomous navigation.In this thesis, using Speeded Up Robust Features (SURF) to identify visual beacon.But in order to accelerate image processing and increase the recognition rate, need to do image preprocessing. Based on the characteristics of every visual beacon, define suitable image preprocessing.
In this thesis, the experiment is executed on our campus. According to the environmental characteristics of the campus, set out distinctive visual beacons. Advance to detect visual beacons' interest points and descriptors, and store to pad's memory to facilitate the subsequent alignment use. And save time to detect interest points and extract their descriptors repeatedly. It's impotant to select right region of interest (ROI) . First, in the area of visual beacon that may appear in the image, select a wide range of the region of interest. Second, if visual beacon has significant visual color characteristics, and then detect the color of a real region of interest. This not only speeds up the image processing speed and increases the recognition rate. Further, there are many trees on our campus and the trees are thick with leaves. There is a large number of interest points come from those leaves. Need to spend a lot of time to compute. However, these interest points are not fixed, and can not be used. Therefore, this paper through the color detection to remove pixles of the leaves. Not only saves a lot of time in operation to detect interest points and extract descriptors, but also to enhance the recognition rate.

[1] M. A. G. Moreira, H. N. Machado, C. F. C. Mendonc﹐G. A. S. Pereira,“Mobile robot outdoor localization using planar beacons and visual improved odometry,”Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 2468–2473, 2007.
[2] M. Hamada, A. Boutarfa,“A New Mmethod Using Hough Transform (HT) in Robotic Vision,”Journal of Electrical Engineering, vol. 14, pp. 8-17, 2014.
[3] H. Bay, T. Tuytelaars, L. V. Gool, “SURF: Speeded up robust features," 9th European Conference on Computer Vision, pp. 404-417, May 2006.
[4] T. Gevers, A. W. M. Smeulders,“Color-based object recognition,”Pattern Recognition, vol. 32, pp. 453-464, Mar. 1999.
[5] E. Todt, C. Torras,“Color-contrast landmark detection and encoding in outdoor images,”Computer Analysis of Images and Patterns, vol. 3691, pp. 612-619, Sep. 2005.
[6] 沈予平，「基於擴展型卡爾曼濾波器之中型雙足人形機器人足球影像追蹤與定位」，碩士學位論文，國立台灣科技大學，2015年。
[7] D. Gossow, J. Pellenz, D. Paulus,“Danger Sign Detection Using Color Histograms and SURF Matching,”IEEE International Workshop on Safety, Security and Rescue Robotics, Oct. 2008.
[8] 張哲豪，「基於立體視覺之移動式機械臂影像伺服設計」，碩士學位論文，國立交通大學，2010年。
[9] 林家儒，「運用顏色與SURF關鍵點特徵分類之快速多廣告看板計次系統」，碩士學位論文，國立臺北科技大學，2012年。
[10] D. G. Lowe, “Distinctive Image Features from Scale-invariant Keypoints,”International Journal of Computer Vision, vol. 60, pp.91-110, Jan. 2004.
[11] C. H. Kuo, H. C. Chou, S. Y. Tasi,“Pneumatic Sensor: A Complete Coverage Improvement Approach for Robotic Cleaners,” IEEE Transactions on Instrumentation and Measurement, vol. 60, pp. 1237-1256, Apr. 2011.
[12] Y. Aniroh,“Adaptive Gain Sliding Control Based Trajectory Tracking For Wheeled Wall Climbing Robots,”Master's Thesis, National Taiwan Univesity of Science and Technology, Jan. 2014.
[13] K. Rana, M. Zaveri, “A-Star Algorithm for Energy Efficient Routing in Wireless Sensor Network,”Trends in Networks and Communications, vol. 197, pp. 232-241, Jul. 2011.
[14] P. E. Hart, N. J. Nilsson, B. Raphael, “A formal basis for the heuristic determination of minimum cost paths,” IEEE Transactions on Systems Science and Cybernetics, vol. 4, pp. 100-107, Jul. 1968.
[15] C. P. Yadav, R. K. Jain and S. K. Yadav, “An Efficient Routing Method for Lifetime Enhancement in Wireless Sensor Network using Fuzzy Approach and A-Star Algorithm,” International Journal of Engineering and Innovative Technology (IJEIT), Mar. 2014.
[16] V. Benoît, T. Nakashima, “A New Approach to Path-finding by Possibilities Search,” Soft Computing and Intelligent Systems, 2014 Joint 7th International Conference on and Advanced Intelligent Systems, Dec. 2014.
[17] E. W. Dijkstra,“A Note on Two Problems in Connexion with Graphs,”Numerische Mathematik, vol. 1, pp. 269-271, Dec. 1959.
[18] 張書豪，「具自主導航與避障之嵌入式移動機器人控制系統開發」，碩士學位論文，國立臺灣科技大學，2010年。
[19] 戚守為，「考量陰影干擾之自主人形機器人於Robocup足球賽避障導航控制」，碩士學位論文，國立台灣科技大學，2013年。
[20] 李仲凱，「針對FIRA HuroCup自主避障競賽所開發之嵌入式人形機器人控制系統」，碩士學位論文，國立臺灣科技大學，2014年。
[21] 林思羽，「基於Android之FIRA HuroCup自主視覺避障人型機器人應用程式開發」，碩士學位論文，國立臺灣科技大學，2015年。
[22] C. S. Chiu, T. S. Chiang, Y. T. Ye,“Fuzzy obstacle avoidance control of a two-wheeled mobile robot,”International Automatic Control Conference (CACS), Nov. 2015.
[23] G. Dudek, M. Jenkin, Computational Principles of Mobile Robotics, Cambridge University Press, 2nd edition, 2010.
[24] Great-circle distance, access via Wikipedia website (https://en.wikipedia. org/wiki/Great-circle_distance).
[25] Length of latitude, access via Wikipedia website (https://en.wikipedia.org/ wiki/ Latitude).
[26] Length of longitude access via Wikipedia website (https://en.wikipedia.org/ wiki/ Longitude).