---

The Piero sports graphics system, that uses MATRIS line-tracking technology developed by the BBC, won two awards at IBC in Amsterdam, September 2006. Piero won the IBC Innovation Award for Innovative Application of Technology in Content Creation as well as the Cable and Satellite Product of the Year Award for Best Outside Broadcast Technology or Service

---

Red Bee Media, who licensed the MATRIS line-based tracking technology from the BBC, have sold a Piero sports graphics system to Sky Italia. Piero uses the MATRIS technology to calculate the position and orientation of the camera in real time, by tracking the pitch lines, so that virtual graphics can be rendered so as to appear tied to the pitch. Sky Italia's first use of the system was in Milan on 18th April 2006, and they will use it for their coverage of the World Cup. See www.redbeemedia.com for more details.

---

BBC uses technology from European R&D project for enhanced sports coverage.

---

• 
• (a) MATRIS hardware prototype

The MATRIS hardware prototype features a color firewire camera integrated with a high-accuracy miniature inertial measurement unit (IMU). The IMU's low-power digital signal processor pre-processes sensor readings at very high update rates and takes care of exact synchronization with the image grabber, hence supplying a synchronized stream of video, 3D orientation, 3D linear acceleration, angular velocity, and earth magnetic field data. Max update rates are firewire 30 fps color images 640x480, together with 512 Hz inertial data using USB interface.

---

In the offline stage, the input images are processed to build up a representation of the scene, which can be used for tracking in the online phase. First, corresponding interest poinst are searched in the input images (fig. a).

From these correspondences the camera poses and the 3D points referring to the correspondences are estimated (fig. b), resulting in a sparse 3D point cloud.

Having obtained all camera poses, dense depth maps (fig. c) are computed for all images, they contain the spatial depth for each pixel in the image.

These depth maps may be fused into a scene model (fig. d). Either the scene model (fig. d) or the depth maps (fig. c) can be used to select and describe interesting 3D regions which can be used for online tracking.

---

One major goal of MATRIS is to combine the measurements of the inertial sensor MT9 (see figure (a)) with the vision-based tracking data.

The accuracy and stability of the sensor has first to be estimated and modelled in order to enable an optimal fusion. We attached the sensor to the TV-camera, and tracked it with the highly precise BBC-tracking system "free-d" and the sensor-itself (see figure b)). The comparison of the data enabled to confirm the validity of the approach and gave very good estimate for the future modelling of possible motion models and error boundaries."