CentroidPeaksMD: Algorithm to center peaks in reciprocal space

[Janik Zikovsky]

(In [12218]) Refs #3122: Methods for finding the centroid of a sphere. Algorithm started.

[Janik Zikovsky]

(In [12225]) Refs #3122: Algorithm to centroid peaks. Fixed FakeMDEventData giving non-uniform peaks.

[Janik Zikovsky]

(In [12243]) Refs #3102: test for GetEulerAngle(), fixes to universalGoniometers. Also refs #3122: Cxxtest got confused by a commented-out performance test.

[Janik Zikovsky]

(In [12275]) Refs #3122: Centroid peaks in reciprocal space in either Q or HKL.

[Janik Zikovsky]

Join together the integrateSphere method and centroidSphere() so that one method does both.

[Janik Zikovsky]

Moved my tickets that won't get finished by iteration 29

[Janik Zikovsky]

In [13696]:

Refs #3122: Added compiler directive to avoid caching centroids for all boxes - should save some memory at the cost of slower centroiding/peak finding.

[Janik Zikovsky]

This method looks pretty reasonably well working. Some tweaking might be done with real data (e.g. an iterative routine of UB matrix finding, centroiding, etc.) , but will leave it alone for now.

[Janik Zikovsky]

Algorithm renamed to CentroidPeaksMD.

[Peter Parker]

Unsure as how to proceed with testing. Have used CreateMDWorkspace, FakeMDEventData and CreatePeakWorkspace via the UI to create necessary objects to feed into CentroidPeaksMD, but unsure of exactly what I'm looking for.

Perhaps this is best left for someone less new to the project.

[Stuart Campbell]

Janik, how do you test this ?

[Janik Zikovsky]

Here is a sample script using real data:

# Basic parameters
filename = "TOPAZ_3131_event.nxs"
#Name of the workspaces to create
ws = "TOPAZ_3131"

# Load the original data
LoadEventNexus(Filename=filename,OutputWorkspace=ws)

# Convert to reciprocal space, in the sample frame
ConvertToDiffractionMDWorkspace(InputWorkspace=ws,OutputWorkspace=ws+'_MD',
		OutputDimensions='Q (sample frame)',LorentzCorrection='1')
		
# Find peaks
FindPeaksMD(InputWorkspace=ws+'_MD',MaxPeaks='50',OutputWorkspace=ws+'_peaks')

# Find the UB matrix using the peaks and known lattice parameters
FindUBUsingLatticeParameters(PeaksWorkspace=ws+'_peaks',a='10.3522',b='6.0768',c='4.7276',
		alpha='90',beta='90',gamma='90', NumInitial='20', Tolerance='0.12')
# And index to HKL		
IndexPeaks(PeaksWorkspace=ws+'_peaks', Tolerance='0.12')
		
# Centroid the peaks
CentroidPeaksMD(InputWorkspace=ws+'_MD',CoordinatesToUse='Q (sample frame)',
		PeakRadius='0.05',PeaksWorkspace=ws+'_peaks',OutputWorkspace=ws+'_peaks_centered')
		

Alternatively, you can use FakeMDEventData to create fake peaks in an empty workspace and try to get their centroid. You'll need to create a peaks workspace to start at, I'm not sure how much of Peak creation is exposed to Python though... Also, the easiest way to do it that way is to do it in HKL space (otherwise the Peaks need to have a physical meaning in Q and they are harder to fake :) )

[Janik Zikovsky]

Maybe use LoadIsawPeaks to make the initial peaksWS?

[Stuart Campbell]

This ticket has been transferred to github issue ​3969