Add a 'geometry tolerance' variable for pre-compiled applications

[MarkGardnerUSyd]

I create geometry files using Matlab and then run the pre-compiled rtk applications (rtkfdk.exe, rtkadmmtotalvariation.exe etc) on a windows terminal (yes I know it's weird). The use of two different operating systems causes small variations in the geometry file (), giving me errors such as:

File: C:\RTK-2.3.0\RTK-2.3.0\src\rtkThreeDCircularProjectionGeometryXMLFileReader.cxx
Line: 162
Description: ITK ERROR: Matrix and parameters are not consistent.
Read matrix from geometry file: 
-1237.5 -0 862.666 -160000
0 -1500 -0 0
0.48163 -0 0.876374 -1000

Computed matrix from parameters:
-1237.5 0 862.667 -160000
0 -1500 0 0
0.481631 0 0.876374 -1000

I believe this error is displayed when the difference between the two matrices is larger than a certain value, however this value is smaller than the computation error that occurs between different operating systems.

If you were to add an optional variable in the compiled applications that specifies the maximum allowed difference between the two geometry files, then I could increase this value so that this error no longer appears. An example could be something like:

rtkfdk --geometryTolerance e-5

Hope this makes sense.

[MichaelColonel]

It could be easily done. In example of yours the inconsistency is about 0.001 (value in first row, third column) i would say it's quite significant!

[SimonRit]

Thanks for the report. Can you provide the parameter values which were used to build this projection matrix? It would be good to understand the root cause of this behavior difference between OS.

[MarkGardnerUSyd]

We have a Matlab Script that creates the geometry file:

 StartAngle = -180;           %First projection angle
 EndAngle = 180;
 ProjNum = 895;
 angs = linspace(StartAngle,EndAngle,ProjNum) - 90;
 x_offset = -160;

 for jj = 1:numel(angs)
    G(:,jj) = [angs(jj);0;0;0;0;x_offset;0;SID;SDD];
    Pmats(:,:,jj) = gtop(G(:,jj));
 end

 WriteRTKGeometryXML(Pmats,geoFile);

where the function gtop is:

function [P] = gtop(g)
%Takes physically meaningful parameters g to make projection matrix P
%% the P part

GantryAngle = g(1);OutOfPlaneAngle=g(2);InPlaneAngle=g(3);SourceOffsetX=g(4);SourceOffsetY=g(5);DetOffsetX=g(6);DetOffsetY=g(7);SID=g(8);SDD=g(9);

GA = GantryAngle*((2*pi)/(360));                 
OPA = OutOfPlaneAngle*((2*pi)/(360));              
IPA = InPlaneAngle*((2*pi)/(360)); 
SOX = SourceOffsetX;
SOY = SourceOffsetY;
DOX = DetOffsetX;
DOY = DetOffsetY;

%% 
% R1 = [cos(-IPA),-sin(-IPA),0,0;sin(-IPA),cos(-IPA),0,0;0,0,1,0;0,0,0,1];
% R2 = [1,0,0,0;0,cos(-OPA),-sin(-OPA),0;0,sin(-OPA),cos(-OPA),0;0,0,0,1];
% R3 = [cos(-GA),0,sin(-GA),0;0,1,0,0;-sin(-GA),0,cos(-GA),0;0,0,0,1];
% 
% S1 = [1,0,SOX-DOX;0,1,SOY-DOY;0,0,1];
% S2 = [-SID,0,0,0;0,-SID,0,0;0,0,1,-SID];
% S3 = [1,0,0,-SOX;0,1,0,-SOY;0,0,1,0;0,0,0,1];
%
% P = S1*S2*S3*R1*R2*R3;

P = [   [ - sin(GA)*(cos(OPA)*(DOX - SOX) + SDD*sin(IPA)*sin(OPA)) - SDD*cos(GA)*cos(IPA), sin(OPA)*(DOX - SOX) - SDD*cos(OPA)*sin(IPA),   SDD*cos(IPA)*sin(GA) - cos(GA)*(cos(OPA)*(DOX - SOX) + SDD*sin(IPA)*sin(OPA)), SDD*SOX + SID*(DOX - SOX)];...
        [   SDD*cos(GA)*sin(IPA) - sin(GA)*(cos(OPA)*(DOY - SOY) + SDD*cos(IPA)*sin(OPA)), sin(OPA)*(DOY - SOY) - SDD*cos(IPA)*cos(OPA), - cos(GA)*(cos(OPA)*(DOY - SOY) + SDD*cos(IPA)*sin(OPA)) - SDD*sin(GA)*sin(IPA), SDD*SOY + SID*(DOY - SOY)];...
        [                                                                cos(OPA)*sin(GA),                                    -sin(OPA),                                                                cos(GA)*cos(OPA),                      -SID]];

and where WriteRTKGeometryXML prints the geometry matrices to the xml file

[MichaelColonel]

Could it be that the digits precision in Matlab has low value? It looks like the precision is 6, according to your matrix element data.

[SimonRit]

I agree with @MichaelColonel, this seems to be a problem in your WriteRTKGeometryXML, not an OS-specific issue with RTK.

[SimonRit]

BTW, it's possible to use RTK's python code from Matlab if you need to write an XML file from Matlab. I can share an example if you're interested.

[MarkGardnerUSyd]

If you could share an example of using RTK's python code from Matlab to write an XML file from Matlab that would be very helpful. Thanks

[SimonRit]

Here is an example (hope it works, I cannot test it right now)

itk = py.importlib.import_module('itk');
geometry = itk.RTK.ThreeDCircularProjectionGeometry.New();                                                                                                                                                                               
geometry.AddProjection(geometry, 1000., 1536., 28.)
xmlWriter = itk.RTK.ThreeDCircularProjectionGeometryXMLFileWriter.New();
xmlWriter.SetObject(xmlWriter,geometry);
xmlWriter.SetFilename(xmlWriter,'geometry.xml');
xmlWriter.WriteFile(xmlWriter);

[SimonRit]

Any update on this issue?

[MarkGardnerUSyd]

Sorry for the delay. Had to get the itk-rtk python module working.

The attached code works well. Thanks for helping with this issue.

Mark

[SimonRit]

Thanks for the feedback. I don't see any code attached, can you copy paste it please? Can you close the issue if you consider it solved?

[MarkGardnerUSyd]

Sorry I meant that the code you attached above works.