A different idea for doing point cloud self-calibration

[jackjansen]

The current procedure sort-of works, if done by a competent user.

But here's a completely different approach idea. Pretty much "the other way around": we get the user to stand where we want them to be. Comments, please, and assign back to me after that.

When you have to do a view adjust your point cloud representation is removed, and you become a mannequin. (the mannequin is sized to match your height).

You are in an empty room, with footsteps on the floor where you should place your feet. (and instructions in front of you to tell you to walk to that spot).

Once you are on that spot, a table materialises in front of you, with two indicators where you should place your hands. Once your hands are there the mannequin fades out and your point cloud fades in.

At this point we could either ask the user "is this correct", but possibly we could also compute correctness, by cropping three areas out of the point cloud: the left hand, the right hand and the head, and ensuring that there are enough points in each of these areas.