Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: meas_algorithms
-
Labels:None
-
Story Points:1
-
Epic Link:
-
Sprint:Science Pipelines DM-W16-5
-
Team:Data Release Production
Description
This is a port of HSC-1353 and HSC-1360.
Descriptions from HSC:
We typically get a large number of junk detections around bright objects due to noise fluctuations in the elevated background. We can try to reduce the number of junk detections by adding an additional local background subtraction before object detection. We can then add this back in after detection of footprints and peaks.
I forgot to set the useApprox=True for the background subtraction that runs before footprint and peak detection. This will then use the Chebyshev instead of the spline.
Attachments
Issue Links
- mentioned in
-
Page Loading...
Activity
Your plots seem to indicate that turning it on makes things worse in that there are more junk detections. You ran your test on an individual frame, but I only looked at things on the coadds. This probably makes sense in that you would expect more junk for higher S/N. We probably need to test this more before we decide how/where to turn this on.
I totally agree. I was pretty sure you hadn't tried running this on single frames. It looks to me like there are trade offs to both modes, so further testing is certainly needed before (if/where) turning this feature on.
I made the footprintBackground --> tempLocalBackground name change, pushed a tickets/DM-4821 branch, reran Jenkins (successful) and pushed to master.
And release notes have been updated: thanks! 
Bob Armstrong would you mind having a look at this. I've attached two figures demonstrating the difference between running with this feature off (red crosses) and on (green circles). FYI: This was for processCcd.py runs of visit=1228 ccd=71 of the UDEEP COSMOS field.
The decision of whether to turn it on by default (or in config overrides in camera packages) will be left until the ongoing testing is completed on the HSC side of things.
A Jenkins build was successful.