genghisken / ps1

Pan-STARRS and ATLAS code
3 stars 0 forks source link

YSE duplicate objects #172

Open djones1040 opened 4 years ago

djones1040 commented 4 years ago

Hi Ken, this issue has been raised for ATLAS but don't think I've mentioned it for YSE yet. There are a number of duplicate objects in the database, for example sorting the "good" list by spectral classification shows a number of the duplicated objects at the top of the page:

https://star.pst.qub.ac.uk/sne/ps1yse/psdb/followup/2/?sort=-observation_status

genghisken commented 4 years ago

Ah yes. We have an association radius of 0.5 arcsec for Pan-STARRS, but if the scatter is outside this radius, the data will form a new object, even if subsequent detections drift back towards the the central locus of points. In Pan-STARRS at least the PS names are consistent (within 3") for these objects. I've been working on code in ATLAS to merge objects, but this is a non-trivial process! Take a look at the scatter plots in the duplicates and you'll see what I mean. The first detection is shown in the centre of the scatter plot. subsequent detections within 0.5 arcsec are shown in blue. Neighbour first detections are shown in red, with their associated detections in green. For high SNR detections, assuming it's not moving, the scatter should be tight. For low SNR, this has always been an issue for us. In ATLAS, after the first few detections, I reset the detection nearest to the mean RA and Dec to be the defining detection. I don't do this in Pan-STARRS yet.

genghisken commented 4 years ago

For ATLAS, I have code that checks what's already in the Good List, and the web interface returns a message back to the scanner to ask them to place the duplicate in the attic. For some reason I never got round to implementing this in PS1. Time to fix that!

djones1040 commented 4 years ago

thanks Ken!

genghisken commented 4 years ago

Heads up - I've just deployed some new code to stop promotion of duplicate objects to the Good list if they are within 1 arcsec of something ALREADY in the good list.

You should get a message that looks like this:

Screenshot 2020-10-01 at 00 33 04

It's a new bit of code - so let me know ASAP if you see any problems. Looks like it's behaving OK.