This has been floating around in my mind for a while now, and since the Kinect's just went on sale I'm going to pick one up.... so can these be used to aid the process of modelling real-life cars by collecting a point cloud and 'smoothing' that down to surfaces?
The Kinect (in a single instant) will only get you a view from one point, and car sized turntables would be expensive. So can 'we' track the movement/position of a Kinect as it travels around a large object? I am thinking of a stand somewhat similar to a hospital 'drip' stand with the Kinect mounted sideways at the top, so that it can tilt to look up/down. The operator would push the stand around the object, raise/lower/tilt the Kinect and take a series of 'RGBD' pictures.
So maybe 2 optical mice tracking the motion of the floor would get up traverse/turn information, the Kinect's accelerometer could give the vertical tilt (it's mechanically constrained to one axis) and a linear encode or another mouse could get us vertical height changes.
Once we collect a point cloud of a few million points, does this really help us produce a simple/reduced polygon model? And of course there's also a fair bit of work producing a XML to describe the operation of the car...
Simon
PS. Terraroot's Tuk Tuk (http://i11.photobucket.com/albums/a151/ ... s/turn.jpg) made me post...