Details
-
Type:
Story
-
Status: To Do
-
Resolution: Unresolved
-
Fix Version/s: None
-
Component/s: meas_algorithms
-
Labels:
-
Epic Link:
-
Team:Alert Production
Description
The current base_SdssCentroid_flag does not incorporate as much detail about the "goodness" of the centroid fit as I would like. For example, for a random HSC image, I find a non-zero list of objects with centroid_flag False and interpolated, saturated, and/or edge True:
ipdb> dataRef.dataId
|
DataId(initialdata={'taiObs': '2006-06-02T09:07:28.22', 'extension': 13, 'object': 'D3', 'visit': 850587, 'filter': 'r', 'state': 'p', 'runId': '06AL01', 'tract': 0, 'date': '2006-06-02', 'ccd': 12, 'expTime': 300.181}, tag=set([]))
|
ipdb> cent = src['slot_Centroid_flag']
|
ipdb> sat = src['base_PixelFlags_flag_saturated']
|
ipdb> interp = src['base_PixelFlags_flag_interpolated']
|
ipdb> edge = src['base_PixelFlags_flag_edge']
|
ipdb> (sat & ~cent).sum()
|
52
|
ipdb> (interp & ~cent).sum()
|
173
|
ipdb> (edge & ~cent).sum()
|
5
|
I do not have any cases where the centroid values or sigmas are non-finite and Centroid_flag is False, but those were claimed to have been found in some CFHT data in the past.
I would propose that a Centroid_isReliable (or similar name) flag be defined as follows, with fields removed if they are already incorporated into Centroid_flag.
all(
|
not centroidFlag,
|
isfinite(centroid),
|
isfinite(sigma),
|
ok covariance (somehow defined),
|
not saturated,
|
not interpolatedCenter,
|
not edge
|
)
|
Attachments
Issue Links
- mentioned in
-
Page Loading...
Reviewed for DM-CCB, 2020-06-02.
And it made me wonder somewhat about the semantics of flags, but not enough to distract from other work that's currently in progress.