3 ms·
3D scanning with a plain webcam
- jacquesm 17y agoVery impressive, it won't be long or people will be using their webcams to scan 3D avatars of themselves. "The Street" isn't that far off. the original article: http://www.fabbaloo.com/2009/11/real-time-3d-scanning-with-webcam.html http://www.fabbaloo.com/2009/11/real-time-3d-scanning-with-w... Homepage of the guy that made it: http://mi.eng.cam.ac.uk/~qp202/ http://mi.eng.cam.ac.uk/~qp202/
- liuliu 17y agoWell, hard to say. He uses FAST pointer detector for point tracking which may not be this accurate for other "smoothing" surface object e.g. your mobile phone or "sharp-vertices" object e.g. the doll. It seems only can get impressive result for objects with simple 3d structure and detailed texture surface (has good point cloud prediction).
- jacquesm 17y agoIn the presentation there is a good bit of data about feature tracking, not point tracking, I'm not sure how well it would work but this is the best demo I've seen so far of a program making a 3D model without stereoscopic vision and preparation of the subject. What you could add is contour detection (the contours could be added as a constraint on the outer limit of the volume bounded by the surface being scanned). That way you could make a pretty good stab at what the surface must have been like. Contours will always be 90 degrees from the 'normal' of the object being scanned. That trick will not work for concave surfaces though.
- RevRal 17y agoI tried to create something like this a long time ago, but I really wanted it to work by using pictures from only three equidistant angles. Probably why I never finished it....