近年來，隨著特徵匹配技術在衆多領域不斷地被應用，讓影像特徵匹配演算法受到了視覺研究者們越來越多的關注。基於特徵的匹配策略是將同一景物在不同視角下得到的兩幅影像中的同一位置聯繫起來的一個過程，一般通過尋找兩幅影像之間的局部映射關係來完成。主要包括點匹配、線匹配和區域匹配等。到目前為止，關於點匹配的研究越來越多也日趨成熟。大多數的點特徵匹配方法都是利用紋理資訊來進行，因爲基於灰度資訊的影像匹配方法已經相對成熟，而且對於紋理資訊較爲複雜的物體計算量較大，不能有很好的適應性，所以更多的關注開始集中到使用影像特徵幾何關係作爲匹配重點的研究。
本文基於對印刷電路板（PCB）影像紋理特徵的觀察和研究，提出了一種利用影像最大輪廓特徵角點幾何特性（仿射不變量CR）的點匹配方法。因爲印刷電路板（PCB）在組裝綫移動的過程中很容易發生位置的偏移，所以爲了後續相機的矯正，在無法進行人工標記的情況下，我們至少需要四對以上的能夠匹配的自然特徵點對進行目前實際位置的偵測和後續的組裝位置的修正，對印刷電路板（PCB）類型的影像特徵進行研究後，得到一種基於最大輪廓多邊形角點提取結合該角點與周圍角點的幾何關係的點匹配演算法。該演算法對印刷電路板（PCB）影像的匹配結果適應良好，同時其他物體的影像用該方法測試也得到了很好的結果。
關鍵字：角點匹配；幾何特性；仿射不變量；最大輪廓多邊形；CR

In recent years, with the continuous application of feature matching technology in many fields, the image feature matching algorithm has attracted more and more attention from vision researchers.The feature matching strategy is a process of linking the same position in two images of the same scene from different perspectives.It is generally done by finding the local mapping relationship between two images.These, mainly include, point matching, line matching and area matching. So far, the research on point matching has become reached, a mature status. Most point feature matching methods use texture information, because the image matching method based on gray-scale information is relatively mature, and it has a large amount of calculation for objects with more complex texture information and cannot have a good adaptability. Therefore, the use of image feature geometric relations becomes a new focus of matching research.
The new algorithm proposed in this thesis, is based on the observation and research of the image texture characteristics of PCB board, which is a point matching method using the geometric characteristics of the corner points (affine invariant CR) of the largest contour feature of the image. The Printed Circuit Board (PCB) when stops at each production cell is prone to position errors during the assembly process. For implementing the camera correction,we will need at least four pairs of natural feature points that can be matched for subsequent correction if manual marking is not possible. Therefore, the characteristics of correspondingPrinted Circuit Board (PCB) images are studied, and a new point matching algorithm based on the extraction of the corner points of the maximum contour polygon and the geometric relationship between the corner points and the surrounding corner points is proposed. The matching results on the Printed Circuit Board (PCB) images are promising while testing on other types of images also obtained good results.
Keywords：corner point matching; geometric characteristics;affine invariant;maximum contour polygon; CR

[1] Barnes, Nick. (2012). The role of computer vision in prosthetic vision. Image and Vision Computing. 30. 478–479. 10.1016/j.imavis.2012.05.007.
[2] Lowe, D.G. Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision. 60, 91–110 (2004). https://doi.org/10.1023/B:VISI.0000029664.99615.94.
[3] H．Bay，A．Ess，T．Tuytelaars. Speeded-up robust features (SURF)．Computer Vision and Image Understanding，2008，ii0(3)：346—359．
[4] Rublee E, Rabaud V, Konolige K. ORB: An efficient alternative to SIFT or SURF. In Proceedings of IEEE International Conference on Computer Vision, Los Alamitos: IEEE Computer Society Press, 2011: 2564-2571
[5] Rosten E., Drummond T. (2006) Machine Learning for High-Speed Corner Detection. In: Leonardis A., Bischof H., Pinz A. (eds) Computer Vision – ECCV 2006. ECCV 2006. Lecture Notes in Computer Science, vol 3951. Springer, Berlin, Heidelberg. https://doi.org/10.1007/11744023_34
[6] Zhao Qianqian, “Line Matching Method Based on a New Geometric Invariant： CHR IMP Method” [Online]. Available: https://xueweilunwen.com/doc/ 2014588. [Accessed August 2021].
[7] Jia, Qi & Fan, Xin & Gao, Xinkai & Yu, Meiyu & Li, Haojie & Luo, Zhongxuan. (2018). Line matching based on line-points invariant and local homography. Pattern Recognition. 81. 471-483. 10.1016/j.patcog.2018.03.031.
[8] WANG Wan-liang, JIN Yi-ting, ZHAO Yan-wei and ZHENG Jian-wei. Corner Matching Based on Edge Correlation Distance Constraints. Computer Science, 2013, 40(5): 283-286.
[9] Baidu, “影像預處理,” [Online]. Available: https://baike.baidu.com /item/%E5%9B%BE%E5%83%8F%E9%A2%84%E5%A4%84%E7%90%86/5786684. [Accessed August 2021].
[10]  Brook_icv, “高斯濾波器” [Online]. Available: https://www.cnblogs.com/ wangguchangqing /p/6407717.html. [Accessed August 2021].
[11] Itread01.com, “灰度轉換” [Online]. Available: https://www.itread01.com/ content/1545381550.html. [Accessed August 2021].
[12] Canny, J. "A Computational Approach To Edge Detection"（《一種邊緣檢測的計算方法》）. IEEE Trans. Pattern Analysis and Machine Intelligence. 1986, (8): 679–714.
[13] Baidu，“仿射不變” [Online]. Available: https://baike.baidu.hk/item /%E4%BB%BF%E5%B0%84%E4%B8%8D%E8%AE%8A%E9%87%8F/18878768. [Accessed August 2021].
[14] Ding-Jyh Tsaur, Chi-Feng Tang, Chin-Chi Wu, and Woei Lin. 2005. A threshold-based matching algorithm for photonic clos network switches. In Proceedings of the First international conference on High Performance Computing and Communications (HPCC'05). Springer-Verlag, Berlin, Heidelberg, 166–179. DOI:https://doi.org/10.1007/11557654_22
[15] S．Senda，M．Minoh，I．Katsuo．A fast algorithm for the minimum distance classifier and its application to kanji character recognition．Proceedings of the International Conference on Document Analysis and Recognition，1995，1：283-286．
[16] Bag, Soumen and G. Sanyal. An efficient face recognition approach using PCA and minimum distance classifier. 2011 International Conference on Image Information Processing (2011): 1-6.
[17] J. Heikkila and O. Silven. A four-step camera calibration procedure with implicit image correction. Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 1997, pp. 1106-1112, doi: 10.1109/CVPR.1997.609468.
[18] Z. Zhang. A flexible new technique for camera calibration. In IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 22, no. 11, pp. 1330-1334, Nov. 2000, doi: 10.1109/34.888718.
[19] OpenCV, “Pose Estimation”, [Online]. Available: https://docs.opencv.org/4.5.2 /d7/d53/tutorial_py_pose.html. [Accessed August 2021].
[20] OpenCV, “多邊形擬合曲綫approxPolyDP()函数”, [Online]. Available: https://blog.csdn.net/u013925378/article/details/86075230. [Accessed August 2021].
[21] Richard O. Duda and Peter E. Hart. 1972. Use of the Hough transformation to detect lines and curves in pictures. Commun. ACM 15, 1 (Jan. 1972), 11–15. DOI:https://doi.org/10.1145/361237.361242
[22] Itread01.com, “繪製最小外接矩形、最小外接圓”, [Online]. Available: https://www.itread01.com/content/1547588721.html. [Accessed August 2021].
[23] Dilip K. Prasad, Maylor K. H. Leung, Chai Quek, and Siu-Yeung Cho. 2012. A novel framework for making dominant point detection methods non-parametric. Image Vision Comput. 30, 11 (November, 2012), 843–859. DOI:https://doi.org/10.1016/j.imavis.2012.06.010
[24] Mathsgreat.com, “交比：共線四點的一個射影不變比”, [Online]. Available: http://www.mathsgreat.com/article/article_020.pdf. [Accessed August 2021].
[25] LUO Zhongxuan, LIU Yu. Some New Properties of Algebraic Curves and Hypersurfaces in Projective Space. Journal of Systems Science and Mathematical Sciences, 2009, 29(1): 53-62.