3D vision

[m3d88]

Hi, I want my robot to be able to fetch things.
In order to do that it needs to know how far objects are, which means that it needs to see in 3D.
Does anybody know of the best way to do this:
  - use 2 cameras (like 2 eyes) and re-construct the 3D images from the 2 images ?
  - use one camera and an object sensor to know how far the object is ? In that case I think it'll work as long as objects are on the floor and so is the robot, put it won't work it the object is, say, on a table

Anyway, I'm just checking to see what other people have done to solve this.

Thanks in advance,

Mike.

[Steve Joblin]

check out RoboRealm computer vision software... it is free!  they have lots of techniqes, like determining the Center of Gravity to tell how close/far an object is.

[Half Shell]

using one camera you can still determine object size and htus distance (if you know the size of the object). RoboRealm is easyish to use and free and recommended.

In order to use two cameras to determine curvature of objects and thus depth, keep them at a known distance apart such that the difference in location of pixels that SHOULD be the same can give you, through trigonometry, the depth of the object.