Community Forum

Author Topic: translational scan  (Read 695 times)

cr

  • Newbie
  • *
  • Posts: 4
    • View Profile
translational scan
« on: April 27, 2017, 10:54:54 PM »
Hi there,
I have a question regarding a special setup. We have a translation stage, a fixed line scanner camera and different objects to scan and to model. The object is moved through the field of view of the line scanner and after each scan we rotate the object a bit to scan different perspectives of the object. As we tried to retrieve the 3d model using Agisoft we failed for some reason. After background masking we got a lot of tie points, dense cloud points and meshes, but except the tie points on the images nothing is/can be displayed -  neither the dense cloud nor the mesh. We didn't received any error from the console except 'LM singular'. Does somebody know what we did wrong or how to optimize the setup?
Thanks in advance
« Last Edit: April 27, 2017, 10:58:37 PM by cr »

James

  • Hero Member
  • *****
  • Posts: 532
    • View Profile
Re: translational scan
« Reply #1 on: April 28, 2017, 11:48:54 AM »
I think the main problem is that the images you generate do not fit the pinhole/frame camera model at all. Instead of a principal point near the center of the image you have a 'principal line' running the length of the image perpendicular to the scan line orientation.

I don't see that you will have much luck with this approach, but one improvement would be to rotate the camera (along the scan line axis) to scan a static subject rather than keep the camera static and translate the subject.

Alternatively and more creatively you could keep the camera static and rotate the subject about the camera scan line axis - i.e. imagine the subject on the inside surface of a rotating cylinder, with the camera static in the middle of the cylinder.

With either of the two approaches above you would have to make sure the rotation of either the camera or the 'cylinder' was about the camera's nodal point.

cr

  • Newbie
  • *
  • Posts: 4
    • View Profile
Re: translational scan
« Reply #2 on: April 29, 2017, 01:20:12 PM »
Thank you, good ideas. I also tried to mask everything except the object and it worked quite nicely. Anyway, I like the 'cylinder' approach, because we are thinking about a 2 or 3 axis electronically steered rotation between the tranlational scans. Unfortunately, there is no option to move the cameras or to adapt this to standard turnkey solutions. Again, thank you. I have to think about your remarks.

Cheers, Chris