At GDC 2012 Sony UK has presented its new markerless tracking technology for PS Vita called Magnat. The technique allows to track basically any textured 3D environment.
The algorithm is able to initialize itself without any need of markers or pre-learning phase. It is also able to detect planar surfaces in the cloud of reconstructed 3D points. This allows the user to add 3D objects interactively.
Since those planar surfaces can easily be triangulated and the corresponding real video texture is known, they are able to deform these surfaces and render them for example liquid. See the second video for this nice effect.
While I have no further information about the actual algorithm that is used by Magnat, my first guess would be a variation of the popular Parallel Tracking and Mapping approach presented by Klein in 2007.
The algorithm is able to initialize itself without any need of markers or pre-learning phase. It is also able to detect planar surfaces in the cloud of reconstructed 3D points. This allows the user to add 3D objects interactively.
Since those planar surfaces can easily be triangulated and the corresponding real video texture is known, they are able to deform these surfaces and render them for example liquid. See the second video for this nice effect.
While I have no further information about the actual algorithm that is used by Magnat, my first guess would be a variation of the popular Parallel Tracking and Mapping approach presented by Klein in 2007.
more videos after the break
No comments:
Post a Comment