Tankendo Motion Estimation System with Robustness Against Differences in Color and Size Between Users' Clothes Using 4-Color Markers with Elastic Belts

ABSTRACT

To allow an individual to immediately try to train thrusting and striking motions of Tankendo on his / her own on the spot without changing clothes whenever necessary, we suggests a system, which is capable of capturing and analyzing the postures of thrusting and striking motions of Tankendo in a sagittal plane using up to 4-color markers and a single camera, even though there are differences in color and size between users' clothes and something is placed in the background. First, focusing on the colors of the background and users' clothes in a relatively wide area on a captured image, the system has integrated the procedure that different colors are allocated to the color markers of users, who may approach each other through a thrusting / striking motion, while the same color is allocated to the color markers of users, between whom the distance does not change. Moreover, some color marker may be difficult to visually observe when the user moves and overlaps with something with a relatively wide area in the background. To address this problem, a color different from that of the background is allocated to the color marker in question. This makes it easy to set the thresholds for extracting the markers in image processing even though the hue across the captured image reduces. Second, to prevent the color markers from getting out of original positions due to loosened clothes, the elastic belts with color markers fixed on are attached to e user's body. The result of our system evaluation demonstrated that up to 4-color markers might prevent similar color markers from approaching or occluding each other when the users move, regardless of the mode of thrusting / striking, simply by attaching the color markers to the users' bodies in the suggested procedure, even if the color of the clothes or something in the background would change. Moreover, when the color markers were attached to the user's bodies following the procedure doe attaching he markers specified in the system, the articular angle estimation could be achieved with relatively satisfactorily accuracy.

References

1. M.Nakamura and K.Hoshino. 2018. Experimental Discussion about the Influence of the proficiency in Tankendo on the Foot Posture. Proc. 20th Intl. Conf. on Business, Economics, Social Science & Humanities (BESSH-2018), 20, 1--8.
2. K.Hoshino and C.Cheng. 2016. Development of a swordsmanship machine enabling the inner and outer muscles to be safely trained while having fun. Proc. 3rd Intl. Conf. Biomedical and Bioinformatics Engineering. 59--62.
3. Kiyoshi Hoshino, Junya Iwasaki, Maki Nakamura. 2018. Portable system for measurement of martial arts movement, The Institute of Electronics. Information and Communication Engineers (IEICE) Technical Report. 53--56.
4. Kiyoshi Hoshino. 2018. The method for analyzing the motions of martial arts, Patent Application No. 2018--27366, Filling date: Feb 19 2018.
5. Yoshimori Kiriyama, Takeo Nagura, Thomas P.Andriacchi. 2014. Measurement Technique for Ergonomics-Measurement of Body Motion-Motion Measurement and Analysis using Capture System. The Japanese journal of ergonomics, 50, 4, 172--181.
6. Masanobu Yamamoto. 2002. Computer Vision-based Motion Capturing System. IPSJ Transactions on Database.43, 1--9.
7. Takashi Oida, Junichiro Hori, Kazuhiko Takahashii, Masafumi Hashimoto. 2010. Markerless Human Motion Capture Using Multiple Images of 3D Articulated Human CG Model. Transactions of the Japan Society of Mechanical Engineers. 76, 772, 3422--3429.
8. Kojima Taihei, Hiyama Atsushi, Miura Takahiro, Masamitsu Sakai, Michitaka Hirose. 2015. Gait Analysis of the Elderly using VR Gait Visualization System. Proc. The Society of Life Support Engineering, 3E1--06, 1--4.
9. Zhe Cao, Tomas Simon, Shih-En, Wei, Yaser Sheikh, 2017. Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1302--1310. Print ISSN: 1063--6919.