Improve peak fitting capability of FindPeaks

[Wenduo Zhou]

Algorithm CalibrateRectangularDetectors requires a better algorithm to find and fit peaks, especially for POWGEN's K4 detector. Thus algorithm in FindPeaks should be improved to minimize the number of spectra that have peaks fit to wrong positions, which reduces the resolution of summed peaks.

[Wenduo Zhou]

Clean the codes. Refs #7001.

Changeset: aff4398c26e1a23aeffb59635fdcc27fd57f8b09

[Wenduo Zhou]

Add some debug output workspace. Refs #7001.

Changeset: 56ea3a493ec39757020740cac415ef351bcdf7ee

[Wenduo Zhou]

Checkpointing work. Refs #7001.

Changeset: bf8227076d043d74e8603b16f3d6a5c0a7e2c3d1

[Wenduo Zhou]

Checkpointing work. Refs #7001.

Changeset: 3ab5e45cc2c9fb1a342cf3102352cf5921c53f2c

[Wenduo Zhou]

Make some change with FindPeaks. Refs #7001.

Changeset: fe43cec0ae3c1c4f5292a0ac96948dc11f3c5b51

[Wenduo Zhou]

Improve fitting. Refs #7001.

Changeset: a9d4cf97ad218cc8bd5f27d50469cbfe7037b836

[Wenduo Zhou]

Improve FindPeaks. Refs #7001.

Changeset: 227f891073c34a00839e26628babe6a4f3cb3a6e

[Wenduo Zhou]

Corrected a major error. Refs #7001.

The algorithm to handle FitWindows were based on a wrong definition. The code is corrected such that FitWindows specify the range data to fit a peak.

Changeset: 555124239092b94ec8e1920290938af3e3414704

[Wenduo Zhou]

Fixed a bug. Refs #7001.

Changeset: 92960c50123765ebf47e286e297afecc2a1fd459

[Wenduo Zhou]

Improved consistency between 2 algorithms. Refs #7001.

Changeset: 86e6cd7080352daeb4b6752553051fa1129f0f9e

[Wenduo Zhou]

Ignore super wide peaks. Refs #7001.

Changeset: 2104bf18723c6e1c43035f42298c99b0c5daa9d3

[Wenduo Zhou]

Added specific support for flat background. Refs #7001.

Changeset: c757b369f7d9199e1bcca3eca5d99101c3d46828

[Wenduo Zhou]

Add property minimum peak height. Refs #7001.

Changeset: fad64869fa15019ffd3466d825eb4d4bf39006f3

[Wenduo Zhou]

New golden files due to change of FindPeaks. Refs #7001.

Changeset: 5f940db7494c090f036950e6bca6801e1b819118

[Wenduo Zhou]

Tried to resolve a potential conflict in header. Refs #7001.

Changeset: 5f4eeed66a004123014e6b93ea00a8d9fc2308ee

[Wenduo Zhou]

Cleared all the debugging codes. Refs #7001.

Changeset: 64566bf37b8a67c8945bbdd5931aba7cb29a6f4e

[Wenduo Zhou]

Tried to resolve conflict. Refs #7001.

Changeset: 9bfce63f03c600a145db2889e1bd8345c3152f31

[Wenduo Zhou]

Removed some debugging codes. Refs #7001.

Changeset: c761afc1bef90c9a646ef05dfd8bad30afea0b1a

[Wenduo Zhou]

Changed unit test value. Refs #7001.

Changed unit test value as new FindPeaks gives out better result.

Changeset: e8e871acb2aeca8f0b1f10829727b48496d2c88f

[Wenduo Zhou]

Removed optional output property 'OutputWorkspace'. Refs #7001.

Changeset: c4631a05a11df2059c1a8e521fc051cf9f023011

[Wenduo Zhou]

Fixed unit test for FindPeaks. Refs #7001.

Changeset: c4add447b74001e53a25100384f1b0d80896c3f9

[Wenduo Zhou]

Cleaned the doxygen. Refs #7001.

Changeset: 5f08e837bc4e782268dd7b0c6614f149ab3ebe59

[Wenduo Zhou]

Removed some debug output. Refs #7001.

Changeset: e304c143a0c53656d035c8542ad6e7693dec1a3b

[Wenduo Zhou]

Instruction for Verification

The work to solve the issue reported by this ticket turned out to be so complicated that it affects more than one algorithm in Mantid. The reasons include the incapable algorithm and exiting defects that were not found before. Here are the algorithms that have been improved

• FindPeaks
• GeneratePeaks
• CalibrateRectangularDetectors

Suggestion: As the goal of this ticket is to improve performance of FindPeaks, thus the best way to verify the ticket is to compare some results obtained by the original FindPeaks() and modified FindPeaks() commit in this ticket. Please read carefully the instruction for both scripts introduced below before merging the commits of this code to master.

Script 1 to verify the ticket

This script is to show that the new FindPeaks() improves the rectangular calibration.

1. Download script s1A.py
2. Locate data from systemtests/Data/PG3_2538_event.nxs
3. Compile Mantid in master without merging the change from this ticket;
4. Run s1A.py and save workspace PG3_2538_calibrated to a Nexus file;
5. Merge the commits under this ticket to master and build Mantid
6. Run s1A.py again
7. Load previously saved Nexus file and compare to PG3_2538_calibrated in color fill plot. You can see the improvement.

Script 2 to verify the ticket

This script is to show that by the enhanced FindPeaks(), the fitting of Vanadium peaks is improved.

1. Locate the data from system test
   • PG3_4866_event.nxs
   • PG3_FERNS_d4832_2011_08_24.cal
2. Before merging the code, build Mantid and run the attached script s2A.py;
3. Save the output table workspace "FoundPeakTable" to a Nexus file;
4. Merge the commit of this ticket and build Mantid.
5. Run script s2A.py again.
6. Load the previously saved table workspace nexus file;
7. Run GeneratePeaks() on both table workspaces for comparison;
8. Compare the Vanadium peaks generated from the output of original and new FindPeaks. The new vanadium peaks generated by new FindPeaks are more realistic, while some of the peaks found and fit by the old FindPeaks were fit to noise. If you plot the peaks' centre against the their width (sigma), you can see the curve from the new FindPeaks() is more smooth.

[Alex Buts]

I was able to build the ticket and first part of the test looks fine but there are no files needed to test second script available on GIT. And apparently transferring them from SNS is too much pain.

Another one for "SNS only" testing.

[Peter Peterson]

Merge remote branch 'origin/feature/7001_fit_peak'

Re #7001. Conflicts:

Code/Mantid/Framework/Algorithms/src/GeneratePeaks.cpp Code/Mantid/Framework/Algorithms/test/FindPeaksTest.h

Changeset: b91368d53955ea969e6949ab0f331fc6861be1c2

[Stuart Campbell]

This ticket has been transferred to github issue ​7847