Detector-space PeakFinding

[Owen Arnold]

Pascal has developed a detector-space peak-finding algorithm to work on Workspace2D objects. This will need some work to get into Mantid. Need to adjust the output workspace types, add tests etc.

[Owen Arnold]

refs #4222. Change output workspace type.

Problems with Qvector generated Changeset: b15043114cc62889542fbc491a7198d1c481403d

[Owen Arnold]

refs #4222 test original work

Changeset: 5a5e935aea345d73a06d384b3775fb90f7a1de12

[Janik Zikovsky]

Revert "refs #4222 test original work"

This reverts commit 5a5e935aea345d73a06d384b3775fb90f7a1de12.

Changeset: 303846ce1dfcb6e416c8c93d528d6d3942b38737

[Janik Zikovsky]

Refs #4222 just indentation fix

Changeset: 86ab307e61c3473ca1554a1dd9637b2ec40bd675

[Janik Zikovsky]

Refs #4222 fix FindSXPeaks to use our convention for Q

Which is ki - kf, and no 2*pi factor

Changeset: d81c0fb21751fb2fd956e5b6b2503fb8be726c8c

[Owen Arnold]

refs #4222. Cleanup algorithm and add tests.

Changeset: 869015589811ba4b6815a67a339aa490c0e0efba

[Owen Arnold]

refs #4222. Fix GCC Errors and warnings.

Changeset: dd0b10db9a46394954a3659c7f1301c9387ae0ab

[Owen Arnold]

refs #4222. Fix more GCC Errors and warnings.

Changeset: 4a5bf338c5df35f2a4ce860c46751644cbd107bf

[Owen Arnold]

Still have more work to do.

• Move to Crystal directory as suggested by Janik
• Clean up existing code now that we have a good set of tests in place

[Owen Arnold]

refs #4222. Fix type conversion warnings

Changeset: 4549683580a41100c8d625e3888daa7ae6a53a1c

[Owen Arnold]

refs #4222. Fix bug in last commit. Fix warnings

Changeset: f7ad53fcc8b87750dc5b40be55b6b0fa8675d49c

[Owen Arnold]

refs #4222 Move Algorithm to Crystal

Changeset: b940cd93c4a951ab13fbda96aba6498bd6e7f023

[Owen Arnold]

refs #4222 WIKI docs

Changeset: bc9f761729dda8fd4bacef152a1cdf43efb8db80

[Owen Arnold]

refs #4222 Add performance tests

Changeset: 3be1aaa6b6c5483e63a8106b0554d8b34e2b8e0a

[Owen Arnold]

refs #4222 Parallelise FindSXPeaks

Changeset: 6243acef9b728b318fd52b729d8132c90eb12292

[Owen Arnold]

refs #4222 Minor changes to algorithm + better docs

Changeset: 32d4ee958c9b4cce4980126c780452e8fc172b2f

[Owen Arnold]

refs #4222. Fix test error.

Changeset: 960f7a2d2e2414881d01f9333bac12c9940352bf

[Owen Arnold]

refs #4222. Fix test spurious test results.

Changeset: 57c952d318e45faeb61ab95985397a8a503b8572

[Russell Taylor]

Sidestep known Intel compiler idiosyncrasy. Re #4222.

...that causes the compiler to hang indefinitely.

Changeset: 8a2b80f41ffda25ed258e707296fa8a192b3b911

[Owen Arnold]

refs #4222 Fix for 1 bin.

Changeset: b7f83627e1ceb2182d9881ad73eadf927c72d349

[Owen Arnold]

This has been tested out by one of your users and seems to be working okay.

Tester: Please contact me for workspaces since there are specific ones to use for testing. The algorithm seems to work best after running SmoothNeighbours first.

[Owen Arnold]

Test parameters:

RadiusUnits to NumberOfPixels Radius to 3 NumberOfNeighbours to 35 WeightedSum to Flat

See http://www.mantidproject.org/SmoothNeighbours

You can just open up the original peaks workspace in the instrument view and drag the peaks workspace over the top. For ConvertToDiffractionMDWorkspace, Janik could probably provide you with the best parameters.

[Owen Arnold]

Tester:

See attached python script. From that point, you can re-run FindSXPeaks using different background thresholds and tolerance values. I know that for this sample Pascal tends to limit the x range to give better results with this algorithm, but he's not available for me to ask him about this. Although, Pascal and several users have been using this algorithm to date.

[Michael Reuter]

OK, that's better. This looks like it works fine.

[Stuart Campbell]

This ticket has been transferred to github issue ​5069