an initial guess for multi-camera extrinsic calibration

[Hongbo Zhu]

Hi Dr.Pfrommer:

Right now I'm using tagslam for multi-camera extrinsic calibration. Following the test_6 in tagslam_test, I have done it and have an reasonable output.
I'm wondering if it's possible to optimize the extrinsic from an initial guess. For example, I can measure parameters of a 3d model and feed them into tagslam. Optimization will then start from this initial value.

[Bernd Pfrommer]

Use camera_poses.yaml to feed in an initial guess. The R matrix there determines how strongly that initial guess will be weighted. The larger R, the stronger the initial guess will be weighted. Check afterwards that the output camera_poses.yaml has indeed changed, if only a bit.

--
You received this message because you are subscribed to the Google Groups "tagslam" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tagslam+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tagslam/e5902cc9-b0d5-4530-8501-d97da568d86bn%40googlegroups.com.

[Hongbo Zhu]

Thanks! The R matrix works. But I found that there are two output files: camera_poses.yaml and calibration.yaml. Both of them give the pose of each camera, but the values are different. What is the difference between these two outputs?

[Bernd Pfrommer]

From looking at the code, camera_poses.yaml has the extrinsic calibration only whereas calibration.yaml should have both intrinsic and extrinsic calibrations. The extrinsic calibrations should be identical (although camera_poses.yaml also has the R-matrix).