def fit_fault_set(ax, faults, split_axis=1, title='Poles to Planes'):
"""Given a sequence of swfaults representing a bi-modal distribution, fit
a plane to each of the modes."""
vol = populations.vol
faults = list(faults)
numfaults = len(faults)
normals = [norm for fault in faults for norm in normal_vecs(fault, vol)]
numvecs = len(normals)
normals = np.array(normals)
strikes_dips = fit_bimodal_bidirectional(normals, split_axis=split_axis)
strike, dip = smath.vector2pole(*normals.T)
ax.pole(strike, dip, 'ko', markersize=2, alpha=0.5)
cax = ax.density_contourf(strike, dip, sigma=3)
ax.density_contour(strike, dip, linewidths=2, sigma=3)
cbar = ax.figure.colorbar(cax, ax=ax)
cbar.set_label('Number of standard deviations', rotation=-90)
conj = []
for sd, color in zip(strikes_dips, ['r', 'g']):
ax.pole(*sd, color=color)
ax.plane(*sd, color=color, lw=2)
conj.append(u'%0.0f\u00b0/%0.0f\u00b0' % sd)
ax.set_azimuth_ticks([])
# ax.grid(True)
ax.set_title(title + '\n' + ' and '.join(conj), y=1.0, va='bottom')
ax.annotate('{} planes from {} faults'.format(numvecs, numfaults),
xy=(0.5, 0), xytext=(0, -12), xycoords='axes fraction',
textcoords='offset points', ha='center', va='top')
def fit_bimodal(normals, weights=None, k=2):
"""Given a set of vectors, find the resultant vector of two modes using
kmeans to seperate the modes. The vectors will be weighted by "weights",
if specified."""
obs = scipy.cluster.vq.whiten(normals)
codebook, _ = scipy.cluster.vq.kmeans2(obs, k)
mode, _ = scipy.cluster.vq.vq(obs, codebook)
if weights is not None:
weighted_normals = (normals.T * weights).T
else:
weighted_normals = normals
modes = [weighted_normals[mode==i] for i in range(k)]
return (smath.vector2pole(*mode.mean(axis=0)) for mode in modes)
def test_vector2pole(self):
for (x, y, z), (strike, dip) in self.data:
assert np.allclose(smath.vector2pole(x, y, z), [[strike], [dip]])
def test_vector2pole(self):
for (x,y,z), (strike, dip) in self.data:
assert np.allclose(smath.vector2pole(x,y,z), [[strike], [dip]])
