3
$\begingroup$

I am trying to get started with a stereo experiment and wanted to know the mistakes I might have made during the experiment. The goal is to calculate the depth map of an object using one light projector and one camera mounted in a stereo configuration.

I have a DLP 4500 lightcrafter of resolution ‎912x1140 made by Texas Instruments and a camera of resolution 600x800. I am planning to use triangulation methods to estimate the depth. This is how I have setup the components.

1) The experiment is conducted in an enclosure to eliminate effects of ambient light.

2) The structured light pattern that has to be projected onto the object is loaded as an image. The image is projected onto one of the walls of the enclosure. The projector lens is focused until the projection is clear and the projector is moved backwards or forwards until the projection fits within the boundaries of the face of the enclosure that the projector projects on.

3) The camera is mounted on an adjustable stand with all six degrees of freedom. It is adjusted facing the projection. I am unsure about this part-The camera is placed in the same vertical line as the lens of projector. The camera is placed above the projector at a distance of about 15 cm. This would be the baseline b for the triangulation formula.

4) The lens of the camera is adjusted until the image observed in the camera is sharp and the entire object and some part of the projected image is made to appear in the frame of the camera.

5) The camera is calibrated using a checkerboard in ROS. The camera intrinsics matrix is generated and thus the focal length f required for the triangulation formula is obtained.

Is this setup okay for the experiment? Have I made a mistake at some stage. I am new to computer vision and may have overlooked some aspects. Thanks.

$\endgroup$

1 Answer 1

1
$\begingroup$

Sounds good to me but there are some missing stages.

  1. camera to the projector extrinsic calibration
  2. projector intrinsic calibration -> don't need this stage if you are not interested in the accuracy of your result.

Also, you need to make sure to calibrate camera after adjusting focus. Extrinsic as well if you have made a change in camera-to-projector distance or angle.

$\endgroup$
12
  • $\begingroup$ 1. Doesn't the checkerboard calibration calibrate both the extrinsics and the intrinsics? $\endgroup$ Jan 12, 2019 at 7:45
  • $\begingroup$ 2. The display from the projector doesn't appear warped. Doesn't it mean the projector intrinsics have already been accounted for? $\endgroup$ Jan 12, 2019 at 7:57
  • $\begingroup$ extrinsic in the camera calibration is checkerboard to camera. you need to find camera to projector as well. $\endgroup$ Jan 13, 2019 at 0:41
  • $\begingroup$ any projector image should not look distorted. if TI offers the norminal intrinsic values, it will be fine. don't you need to know the projector intirnsic for the trigonometry? $\endgroup$ Jan 13, 2019 at 0:47
  • 1
    $\begingroup$ 2. I am not sure what kind of decoding method you have in mind but this tutorial is generally very helpful. researchgate.net/publication/… $\endgroup$ Jan 14, 2019 at 1:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.