New Live Sports Graphics System Powered by Realtime Object Tracker
To show comprehensible sports scene
We have developed a new graphics system intended for use in live sports broadcasting. The three-dimensional position of the ball needed by the scene analysis of sports is calculated with high precision and in real time, and this system can achieve accurate CG compositions.
- Use of machine learning for robust object tracking
- By using an object-tracking method based on machine learning, we can robustly track the ball under a variety of circumstances. Also, by processing video at a higher frame rate, we can track faster ball movements than before.
- Camera calibration※ technique for multiple movable cameras
- We have developed calibration technique for multiple movable cameras, that can perform processing with high accuracy and in real time. This allows us to perform high-precision 3D position calculations and CG composition in real time, even for fast-moving cameras and subjects.
Along with improving the performance of each constituent technology, we are working towards verifying this system for a variety of sports so that it can be used it at the 2020 Tokyo Olympic and Paralympic Games.
- ※Camera calibration: The calculation of information such as the camera’s position/orientation, angle of view and lens distortion