穴道按摩是中醫治療的主要方式之一。最近，我們實驗室的陳學長發展了一套能利用手機去預估人臉穴道，並將穴道視覺化在手機螢幕上的系統。使用者能利用此套系統，一邊看著螢幕一邊按壓穴道。然而，在某些情況下，像是手或背部，我們無法同時一手拿著手機一手按摩穴道。在此篇論文，我們希望打造一個基於陳學長的穴道定位方法上，加入機械手臂，使原本的手動按摩替換成自動按摩，達成一個自動穴道按摩的系統。由於陳學長的穴道定位方法是基於2D影像，也因現今的深度攝影機仍十分昂貴和不普及，且我們想要打造一個能容易的打造出來又不會昂貴的系統，所以最終我們嘗試使用2D的相機來打造我們的系統。相機與機械手臂為兩個分開獨立的系統，我們必須找出他們之間的轉換關係，而這種轉換關係也被稱為「手眼標定」，我們使用了透視轉換的方式去找出。由於我們所使用的方式及器材的限制，會導致如果穴道的實際高度不在校正時的高度時，會因為視角的關係而產生誤差，為了解決此問題，我們一樣利用了透視轉換的方式去找出他們之間的關係氣校正誤差。最終，我們打造出了一個能讓使用者容易上手操作的系統，使用者不須事先有穴道的專業知識，只需透過手機的虛擬醫生問診，我們的系統就會自動的透過機械手臂幫使用者實施穴道按摩。

Acupressure is one of the main ways of Chinese medicine treatment. Recently, Chen et al. developed a system on a smartphone to estimate and visualize acupoints on a human face. Users can use the position displayed on the phone to massage the acupuncture points directly. However, in some cases, such as hand or back acupoints, users can’t watch the screen and massage acupoints with a hand at the same time. In this paper, based on Chen's acupoint estimation method, we use a robot arm in our system to replace the manual massage to achieve an automatic massage system. Because Chen's acupoint estimation method is built on 2D image, and depth camera is still expensive and unpopular today, and we want to build an easy-to-build and non-costly system, finally, we try to build our system with a 2D camera. Since the camera and the robot arm are two separate systems, we must find out the conversion relationship between them so-called hand-eye calibration. In order to establish a conversion relationship between the robot arm and the camera, we use perspective transformation to achieve hand-eye calibration. Due to the limitations of the methods, principles, and equipment we use, if the actual height does not fall at the height of the calibration, errors will occur. To solve the height error, we also used perspective transformation to correct the error. Finally, we build a portable, easy to operate and user-friendly system, so that users do not have to massage the acupuncture points themselves, and do not need to have the expertise of Chinese medicine first, just talk to the virtual doctor on the smartphone and place the part to be massaged in the massage area, and the system will automatically massage with the robot arm.

[1] CHEN, Yi-Zhang, et al. Localization of Acupoints using Augmented Reality. In: Proceedings of the 8th ACM on Multimedia Systems Conference. ACM, 2017. p. 239-241.
[2] INADA Massage Chair. http://inada.kenk.com.tw/
[3] LIN, Jen-Shiang. 於視訊影像之穴道定位. 成功大學電腦與通信工程研究所學位論文, 2009, 1-37.
[4] GAO, Huanbing, et al. Research and development of Chinese medical massage robot. Jiqiren(Robot), 2011, 33.5: 553-562.
[5] WU, Chia-Yu. 應用於視訊按摩機之背部阿是穴立體定位. 成功大學電腦與通信工程研究所學位論文, 2010, 1-45.
[6] JIANG, Haotian, et al. Acu glass: quantifying acupuncture therapy using Google glass. In: Proceedings of the 10th EAI International Conference on Body Area Networks. ICST (Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering), 2015. p. 7-10.
[7] Menglong Chang and Qing Zhu. Automatic location of facial acupuncture-point based on facial feature points positioning. 5th International Conference on Frontiers of Manufacturing Science and Measuring Technology (FMSMT), 2017, 545-549
[8] ZENG, Hong, et al. Closed-Loop Hybrid Gaze Brain-Machine Interface Based Robotic Arm Control with Augmented Reality Feedback. Frontiers in neurorobotics, 2017, 11: 60.
[9] 万晓峰. 基于 SCARA 机械手的手眼标定. 电子世界, 2017, 6: 76-77.
[10] KAZEMI, Vahid; SULLIVAN, Josephine. One millisecond face alignment with an ensemble of regression trees. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2014. p. 1867-1874.
[11] ZHAO, Yufeng; ZHANG, Dong; WANG, Yaxiang. Automatic location of facial acupuncture-point based on content of infrared thermal image. In: Computer Science and Education (ICCSE), 2010 5th International Conference on. IEEE, 2010. p. 65-68.
[12] GRAÑA, Rafaella, et al. Development of a low-cost robot able to write. In: Electronic Congress (E-CON UNI), 2017. IEEE, 2017. p. 1-4.
[13] ANDHARE, Pratiksha; RAWAT, Sayali. Pick and place industrial robot controller with computer vision. In: Computing Communication Control and automation (ICCUBEA), 2016 International Conference on. IEEE, 2016. p. 1-4.
[14] WANG, Tao; LIU, Zhaotong; LI, Yanping. Research and realization of the Chinese massage robot based on three-dimensional model. In: Intelligent Control and Automation (WCICA), 2010 8th World Congress on. IEEE, 2010. p. 6493-6497.
[15] Open source computer vision library. https://github.com/opencv/opencv
[16] Davis E. King. A toolkit for making real world machine learning and data analysis applications in c++. http://dlib.net
[17] Davis E. King. Dlib-ml: A Machine Learning Toolkit. Journal of Machine Learning Research, 10:1755-1758, 2009
[18] KING, Davis E. Max-margin object detection. arXiv preprint arXiv:1502.00046, 2015.
[19] YINGZI, Wei, et al. A prediction modeling method for the feet massage robot positioning. In: Intelligent Control and Automation (WCICA), 2014 11th World Congress on. IEEE, 2014. p. 6105-6109.
[20] HU, Zhan-Yi; LEI, Cheng; WU, Fu-Chao. A short note on P4P problem. Acta Automatica Sinica, 2001, 27.6: 770-776.
[21] HU, Fengjun. A rapid eye-to-hand coordination method of industrial robots. JOURNAL OF INFORMATION &COMPUTATIONAL SCIENCE, 2013, 10.5: 1489-1496.
[22] NIOLA, V.; ROSSI, C.; SAVINO, S. Perspective transform and vision system for robotic applications. In: Proceedings of the 5th WSEAS International Conference on Signal Processing, Robotics and Automation. World Scientific and Engineering Academy and Society (WSEAS), 2006. p. 87-91.
[23] DINHAM, Mitchell; FANG, Gu. A low cost hand-eye calibration method for arc welding robots. In: Robotics and Biomimetics (ROBIO), 2009 IEEE International Conference on. IEEE, 2009. p. 1889-1893.
[24] MUELLER, Franziska, et al. GANerated Hands for Real-Time 3D Hand Tracking from Monocular RGB–Supplementary Document–.
[25] The Homography transformation http://www.corrmap.com/features/homography_transformation.php
[26] Regard3D http://www.regard3d.org/
[27] LeCun, Y., Bengio, Y. and Hinton, G., 2015. Deep learning. Nature, 521(7553), pp.436-444.