0

My project is currently required to calculate accurate pose for a very near focusing camera (1-3 inches). To do that, we need to know the Z-value of interesting points and triangulate them to 3D space. We try to consider sensors such as PrimeSense 3D depth camera, which will directly give us depth map but most of the sensor's minimum distance is above 30 cm. Is there any way to accurately get Z-value in very close distance?

SimaGuanxing
  • 673
  • 2
  • 10
  • 29
  • What about a calibrated stereo rig with two macro cameras? You can then rectify the images and calculate the depth map. – aledalgrande Apr 30 '15 at 05:31
  • Can you give more details about the problem you are trying to solve? If you have control over the object/problem you can get depth from a single image by simply having an object of known size in the field view of the camera. – bendervader Apr 30 '15 at 06:44
  • @aledalgrande Thanks! But our setting is not allowed to use stereo camera. lol – SimaGuanxing May 01 '15 at 15:29
  • @bendervader Sure. We are using near focusing camera to measure pose for medical tools. So the mounting space on the tool is limited and only one camera is allowed. Also, we don't want to stick "known distance" marker such as chessboard on the skin. That's why I feel the problem is hard to solve. – SimaGuanxing May 01 '15 at 15:37
  • Yes, this is a hard problem. First of all, it is not possible to determine absolute/metric depth from a single image without knowing something about the environment. There are two ways to get depth for your case without invasive hardware (that I can think of) (1) precisely known camera motion, (2) precisely known dimensions of something in the scene. The other option if you can is to make your own primesense. If you can mount some form a small projector (laser/IR pattern) then image it. You would be able to perform "photometric stereo" which produces accurate 3D. – bendervader May 01 '15 at 17:42
  • Have you considered artificially extending the optical path for the depth sensor? If your setup is on a lab's bench, you could use a few mirrors to fold it... – Francesco Callari May 01 '15 at 18:03
  • @FrancescoCallari Good idea. I have never done that. but worth trying – SimaGuanxing May 04 '15 at 20:08

0 Answers0