Issues in data rotation

[Maël Allard]

Hi mtex community,

I'm trying to rotate my ebsd data (pole figures and maps), but doing so my texture indexes (J and  M) are not the same before and after rotation as well as what is plotted in pole figures

(texture indexes being yet consistent with pole figures). Also, in the same script when I create a grain subset by selection on shape and misorientation parameters (GAM,

Equivalent Diameter and Tortuosity) it doesn't always corresponds to the same number of pixels before and after rotation.

Is the way the rotation is done wrong ?

Thanks in advance,

Maël Allard

(Nb: I'm using mtex 5.2.3 in Matlab 2018a)

% Rotate data
ebsd = rotate(ebsd,rotation('axis',zvector,'angle',-90*degree))

% Parameters used to define the subset
% GAM
GAM = accumarray(ebsd('indexed').grainId, KAM(ebsd('indexed'),'threshold',10*degree)./degree, size(ebsd), @nanmean);

% Equivalent diameter
Area = area(grains);
eq_diameter = 2 * sqrt(Area / pi)

% convex hull (envelope of grains)
[ch_perim, ch_Area] = convexhull_measures(grains);
perim = perimeter(grains);

% Tortuosity
tortuosity = perim ./ ch_perim;

[Rüdiger Kilian]

Hi Mael,
could you also add how you'd derive the J-index?
Cheers,
Rüdiger

[Maël Allard]

Hi Rüdiger,

Yes sure, here it is:

odf_plag = calcDensity(ebsd('Plagioclase').orientations,'halfwidth',10*degree)
% J index
Jindex_plag = textureindex(odf_plag)
% M index
Mindex_plag = calcMIndex(odf_plag)



Cheers,

Maël

[Ralf Hielscher]

Hi Maël,

obviously the J and M index should not change when rotating the sample. If this happens for your data set then this seems to be a bug. Could you provide us with a small, representative data set and a short script illustrating the issue.

Ralf.

********************************************************************
Ralf Hielscher                   Tel: +371-531-38556
Fakultät für Mathematik               +371-531-22200 (Sekr.)
Technische Universität Chemnitz  Fax: +371-531-22109
Reichenhainer Str. 39            E-mail: ralf.hi...@mathematik.tu-chemnitz.de
D-09126 Chemnitz                 http://www.tu-chemnitz.de/~rahi
********************************************************************

--
If you want to reduce the number of emails you get through this forum login to https://groups.google.com/forum/?fromgroups=#!forum/mtexmail, click "My membership" and select "Don't send me email updates". You can still get emails on selected topics by staring them.
---
You received this message because you are subscribed to the Google Groups "MTEX" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mtexmail+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mtexmail/71177822-8364-4a46-987d-7577bebc20a7%40googlegroups.com.

[Maël Allard]

Hi Ralf,

Files are in the downloading link:

https://filesender.renater.fr/?s=download&token=29864251-649e-d16d-e3e6-f228887e955b

Thanks again for your responsiveness,

Maël

[Rüdiger Kilian]

Hi,

I think this is indeed a bug and I guess it's related to this one here

https://github.com/mtex-toolbox/mtex/commit/67c069679f056e6f0c21f190b8e624fb9f4d85ed#diff-7d1ae2500b9b591f63fc3d7480607fa4

in geometry/@quaternion/times.m

Undoing it resolves it for me.

Simple test:

%some input

o = discreteSample(unimodalODF(orientation.rand(1,crystalSymmetry('432')),'halfwidth',15*degree),1000);

odf = calcDensity(o,'halfwidth',10*degree);

% rotated orientations and computer odf again

rot=rotation('axis',zvector,'angle',-90*degree);

odfr = calcDensity(rot*o,'halfwidth',10*degree);

J1 = textureindex(odf)

J2 = textureindex(odfr)

Cheers,

Rüdiger

[Rüdiger Kilian]

Sorry, previous message was a bit premature, it not due to this commit it seems.

Rüdiger

Test with '-1' or '1' for example:

odfin = unimodalODF(orientation.rand(1,crystalSymmetry('-1')),'halfwidth',15*degree);

o=discreteSample(odfin,1000)

odf = calcDensity(o,'halfwidth',10*degree);

% rotated orientations and computer odf again

rot=rotation('axis',zvector,'angle',-90*degree);

or = rot.*o;

odfr = calcDensity(or,'halfwidth',10*degree);

[Rüdiger Kilian]

Adding 'exact' in calcDensity seems to be a workaround.

Cheers,

Rüdiger

[Ralf Hielscher]

Hi Mael,

this was indeed an extremely bad bug that effects many computations. I'm very grateful that you reported it. The bug fix is included in the new version MTEX 5.2.6 which I released just now.

Thank you very much.

Ralf.

[Maël Allard]

Hi Ralf and Rüdiger,

Thanks a lot to both of you for having fix this bug !!

Gratfully,

Maël