doHClustering edit page

hierarchical clustering of rotations and vectors

Syntax

[c,center] = doHCluster(ori,'numCluster',n)
[c,center] = doHCluster(ori,'maxAngle',omega)

Input

ori orientation
n number of clusters
omega maximum angle

Output

c list of clusters
center center of the clusters

Example

% generate orientation clustered around 5 centers
cs = crystalSymmetry('m-3m');
center = orientation.rand(5,cs);
odf = unimodalODF(center,'halfwidth',5*degree)
ori = odf.discreteSample(3000);
odf = SO3FunRBF (m-3m → xyz)
 
  multimodal components
  kernel: de la Vallee Poussin, halfwidth 5°
  center: 5 orientations
 
  Bunge Euler angles in degree
     phi1     Phi    phi2  weight
  156.958 109.836 223.608     0.2
  9.33344 126.207 190.491     0.2
  197.878 76.1995 48.4488     0.2
  156.716 113.621 184.888     0.2
  151.332 117.797 66.3984     0.2
% find the clusters and its centers
tic; [c,centerRec] = calcCluster(ori,'method','hierarchical','numCluster',5); toc
Elapsed time is 7.073283 seconds.
% visualize result
oR = fundamentalRegion(cs)
plot(oR)
oR = orientationRegion
 
 crystal symmetry:  432
 max angle: 62.7994°
 face normales: 14
 vertices: 24
hold on
plot(ori,ind2color(c))
caxis([1,5])
plot(center,'MarkerSize',10,'MarkerFaceColor','k','MarkerEdgeColor','k')
plot(centerRec,'MarkerSize',10,'MarkerFaceColor','r','MarkerEdgeColor','k')
hold off
plot 2000 random orientations out of 3000 given orientations
%check the accuracy of the recomputed centers
min(angle_outer(center,centerRec)./degree)
ans =
    0.6614   13.1669    0.3515   16.6213    0.1755