PROJECTS:
Marker-free real-virtual interaction

	A presenter draws a 3D painting in space by moving a coloured ball.

Existing methods to allow interaction between real and virtual elements in a scene use markers or sensors to track the position of either the real or the virtual object.

Examples may include fitting sensors to the human body, or handling a virtual object using the MixTV marker tracking system.

We are investigating a number of non-intrusive methods to allow interaction between real and virtual objects.

Earlier work in this area includes our 2D actor tracking system, which uses a top-down camera in a studio environment to locate the position of objects (including people) on the studio floor. This allows actors to walk in front and behind virtual objects by switching appropriate mask signals or, by measuring speed and direction of a foot, to kick a virtual football rendered on the studio floor.

For situations where 3D interaction is required, we have shown that the volumetric model of a human generated using the ORIGAMI multi-camera system can be used to trigger events in the virtual world. During our 2003 Open Days, we demonstrated that a presenter could trigger a reaction from a computer-generated dinosaur by punching his or her hand towards the virtual creature.

Recently, we have demonstrated a system that by tracking the position of an object of a known colour (e.g. a hand or coloured ball), allows true three-dimensional drawings to be created in space. This system uses two spotter cameras feeding video to two PCs running software chroma-keyers to identify the position of the object and calculate its position in the studio reference frame. Coordinates are fed into a rendering computer, which draws a trace. The 3D virtual painting can then be viewed from a movable camera, whose position may be measured using the free-d tracking system.