Real-time visualization of sword trajectories in fencing matches
- PDF / 3,180,568 Bytes
- 15 Pages / 439.37 x 666.142 pts Page_size
- 49 Downloads / 173 Views
Real-time visualization of sword trajectories in fencing matches Masaki Takahashi 1 & Shinsuke Yokozawa 1 & Hideki Mitsumine 1 & Tetsuya Itsuki 1 & Masato Naoe 1 & Satoshi Funaki 1 Received: 23 April 2019 / Revised: 28 April 2020 / Accepted: 24 June 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
We developed a system called Sword Tracer that visualizes sword trajectories in fencing matches. Sword Tracer tracks the tips of the swords in the image coordinates and visualizes their movements with computer graphics (CGs). It measures each sword’s position in the infrared (IR) image by detecting IR light reflected from retroreflective tape placed on the tip of the sword. It uses only a single camera and a single marker at the tip, so the system is compact enough to be used in official fencing matches. It accurately detects the tips of the swords by using supervised machine learning and tracks them by predicting their positions in the next frame. The trajectory CGs of the sword tips can be composited on the broadcast image in real-time. Sword Tracer was first used for a broadcast at the All Japan Fencing Championships in December 2017 and has since been used for four other broadcast programs and five exhibition events from 2018 to 2020. TV viewers and guests at the events approved of this new video effect because it helped them to follow the fast-moving swords and gain a better understanding of the swordplay. Keywords Visual object tracking . Machine learning . Fencing
1 Introduction This paper describes Sword Tracer [1], a system we developed for visualizing sword trajectories in fencing. Fencing swords are very thin and move so fast that it is difficult for viewers to follow their movements even in slow-motion video replays. To resolve this issue, our system tracks the tips of the swords in the image coordinates and visualizes their movements with computer graphics (CGs).
* Masaki Takahashi takahashi.m–[email protected]
1
Japan Broadcasting Corporation (NHK), 2-1-1, Jinnan, Shibuya-ku, Tokyo, Japan
Multimedia Tools and Applications
Sword Tracer measures each sword’s position in the infrared (IR) image by detecting IR light reflected from retroreflective tape placed on the tip of the sword. Although conventional optical motion capture systems also use IR light [2], they require multiple IR cameras and numerous reflective markers to be placed on the players’ bodies. In contrast, Sword Tracer uses only a single camera and a single marker at the tip of the sword, which makes it compact enough to be used in official fencing matches. Tracking of the sword in a video sequence needs to be both fast and accurate for use in a live fencing broadcast. Conventional tracking methods tend to be too slow to track fast-moving swords [3, 4]. In contrast, Sword Tracer accurately detects the tips of the swords by using supervised machine learning and then tracks them by predicting their positions in the next frame. The system composites the trajectory CGs of the sword tips on the broadcast image in r
Data Loading...
