Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-27921

Investigate shrinking default pixel padding in astrometric reference object loading

    XMLWordPrintable

    Details

    • Type: Story
    • Status: To Do
    • Resolution: Unresolved
    • Fix Version/s: None
    • Component/s: meas_astrom
    • Labels:
      None

      Description

      Work on DM-24024 revealed a bug where the WCS set on icExp images was getting erroneously shifted away from our best-guess value from the raw exposure (which is based on boresight + rotation + cameraGeom (distortion) as of DM-20154). The effect is seen in a shifted (from our actual best-guess) center for the reference object loading sky circle. This issue was fixed in DM-27868. Now, with a better centering of the sky circle within which reference objects are loaded, it may be reasonable to decrease the padding of the bounding box that is set to ensure the actual CCD sky footprint's worth of reference objects are loaded and considered for the astrometric calibration. Not only could this speed things up with the smaller number of sources, cases have been seen where the WCS fit improves with the smaller pixel padding (see comment for an example). The optimal setting will require testing at a fairly large (PDR2?) scale, but the results of DM-24024 will serve as at least a good guide for HSC data and will perhaps provide enough troublesome cases for testing (hence it being labeled as a blocking ticket).

        Attachments

          Issue Links

            Activity

            Hide
            lauren Lauren MacArthur added a comment -

            Thanks for the info John!  That is certainly something to consider for when/in what context (gen2 or gen3) this ticket gets worked on (seems just sticking to a basic exploration of the current two config parameters might be best if the former).

            Show
            lauren Lauren MacArthur added a comment - Thanks for the info John!  That is certainly something to consider for when/in what context (gen2 or gen3) this ticket gets worked on (seems just sticking to a basic exploration of the current two config parameters might be best if the former).
            Hide
            lauren Lauren MacArthur added a comment -

            Good point re high-declination.  PDR2 gives us plenty of data close to Dec = 45 deg, but testing even more extreme coords would be a good idea.

            Show
            lauren Lauren MacArthur added a comment - Good point re high-declination.  PDR2 gives us plenty of data close to Dec = 45 deg , but testing even more extreme coords would be a good idea.
            Hide
            cmorrison Chris Morrison [X] (Inactive) added a comment -

            Dec 45 should hopefully be enough. I was more worried about data being only like <10 or <5.

            Show
            cmorrison Chris Morrison [X] (Inactive) added a comment - Dec 45 should hopefully be enough. I was more worried about data being only like <10 or <5.
            Hide
            yusra Yusra AlSayyad added a comment -

            Lauren MacArthur, what's left to be done on this ticket?

            Show
            yusra Yusra AlSayyad added a comment - Lauren MacArthur , what's left to be done on this ticket?
            Hide
            lauren Lauren MacArthur added a comment -

            So, in principal, one element of this ticket got sort-of implemented on DM-30030 (but we didn't explore going even lower than 250 pixel padding for any particular dataset).  I do think there are some ideas noted in the comments above that are worth exploring when we have the bandwidth to do so (in particular, I'd like to explore the feasibility of culling reference/source catalogs of regions of NO_DATA before sending to the matcher...), but I'll defer to you if we should hold this ticket for that or just WON'T FIX this one and open a new one later if the future world of time available for "things I'd like to explore" ever materializes!

            Show
            lauren Lauren MacArthur added a comment - So, in principal, one element of this ticket got sort-of implemented on DM-30030 (but we didn't explore going even lower than 250 pixel padding for any particular dataset).  I do think there are some ideas noted in the comments above that are worth exploring when we have the bandwidth to do so (in particular, I'd like to explore the feasibility of culling reference/source catalogs of regions of NO_DATA before sending to the matcher...), but I'll defer to you if we should hold this ticket for that or just WON'T FIX this one and open a new one later if the future world of time available for "things I'd like to explore" ever materializes!

              People

              Assignee:
              Unassigned Unassigned
              Reporter:
              lauren Lauren MacArthur
              Watchers:
              Chris Morrison [X] (Inactive), Eli Rykoff, Ian Sullivan, Jim Bosch, John Parejko, Lauren MacArthur, Yusra AlSayyad
              Votes:
              0 Vote for this issue
              Watchers:
              7 Start watching this issue

                Dates

                Created:
                Updated:

                  Jenkins Builds

                  No builds found.