Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-20541

Determine if any existing obs packages have a flip from PIXELS->FOCAL

    Details

    • Story Points:
      1
    • Epic Link:
    • Sprint:
      AP F19-1
    • Team:
      Alert Production

      Description

      To better understand RFC-605, we need to determine if any of our existing obs packages have a flip in their pixel->focal transform. We think they are all defined as looking "down" on the focal plane, but we need to check.

      1. Load the cameraGeom for each obs package.
      2. Compute determinant of pixel->focal transform.
      3. If determinant is negative, there is a flip.

        Attachments

          Activity

          Hide
          Parejkoj John Parejko added a comment -

          Jim Bosch: Please take a look at the attached code. You can run it with lsst_distrib setup, and it should print out the determinant for each of the detectors on each included camera. All are positive definite. I think I got the math right, but can you please check (note the shape of the matrix returned by linearApprox)?

          Show
          Parejkoj John Parejko added a comment - Jim Bosch : Please take a look at the attached code. You can run it with lsst_distrib setup, and it should print out the determinant for each of the detectors on each included camera. All are positive definite. I think I got the math right, but can you please check (note the shape of the matrix returned by linearApprox )?
          Hide
          jbosch Jim Bosch added a comment -

          I was too lazy to try to figure out Mapping::linearApprox's wonky conventions (but I trusted Russell Owen to have done so), and just modified the script to use

          np.linalg.det(trans.getJacobian(point))

          and reran it. That gave me the same result you got (no flips), and that's good enough for me.

          Show
          jbosch Jim Bosch added a comment - I was too lazy to try to figure out Mapping::linearApprox 's wonky conventions (but I trusted Russell Owen to have done so), and just modified the script to use np.linalg.det(trans.getJacobian(point)) and reran it. That gave me the same result you got (no flips), and that's good enough for me.
          Hide
          Parejkoj John Parejko added a comment -

          Thanks. Given that we agree, and there is some attached code we can fall back on if necessary, I think we can close this.

          Show
          Parejkoj John Parejko added a comment - Thanks. Given that we agree, and there is some attached code we can fall back on if necessary, I think we can close this.

            People

            • Assignee:
              Parejkoj John Parejko
              Reporter:
              Parejkoj John Parejko
              Reviewers:
              Jim Bosch
              Watchers:
              Chris Morrison, Jim Bosch, John Parejko, John Swinbank, Lauren MacArthur, Russell Owen
            • Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved:

                Summary Panel