def cluster(parametersobject):
number_of_orientations = parametersobject.parameterdic['Number_of_orientations']
skip = parametersobject.parameterdic['Skip_initial_frames']
data = []
f_framelist = open('analysis/frames_read.txt', 'w+')
framelist = []
for i in range(1, 1+number_of_orientations):
with open('analysis/coord_matrix_'+str(i).zfill(3)+'.txt', 'r') as f:
l = f.readlines()
data.extend([[float(p) for p in line.strip().split()[1:10]] for line in l[skip:]])
f_framelist.write(str(len(l))+'\n')
framelist.append(len(l)-skip)
f_framelist.close()
ms = MeanShift(n_jobs = -2, cluster_all = True)
ms.fit(data)
labels = ms.labels_
names, numbers = np.unique(labels, return_counts = True)
cluster_centers = ms.cluster_centers_
fout = open('analysis/clusters.txt', 'w+')
fout.write('label\tcount\ttheta\tphi\ttheta_x\ttheta_y\ttheta_z\tx\ty\tz\tR[0][0]\tR[0][1]\tR[0][2]\tR[1][0]\tR[1][1]\tR[1][2]\tR[2][0]\tR[2][1]\tR[2][2]\n')
phi = 0
for i, line in enumerate(cluster_centers):
R = [_[:] for _ in [[]]*9]
x, y, z = line[0], line[1], line[2]
R[0] = line[3:6]
R[1] = line[6:9]
R[2] = np.cross(R[0], R[1])
V = Vector(x, y, z)
if V.norm() > 1e-6:
theta = V.angle(Vector(0,0,1))
norm = np.sqrt(x*x + y*y)
if norm > 1e-6:
phi = np.arctan2(y,x)
#otherwise phi isn't updated and the previous value is copied. Keeps it from jumping near the poles.
else:
theta = 0.0
theta_x = np.arctan2(R[2][1], R[2][2])
theta_y = np.arctan2(-R[2][0], np.sqrt(R[2][1]*R[2][1]+R[2][2]*R[2][2]))
theta_z = np.arctan2(R[1][0], R[0][0])
fout.write(str(names[i])+'\t'+str(numbers[i])+'\t')
fout.write(str(theta)+'\t'+str(phi)+'\t')
fout.write(str(theta_x)+'\t'+str(theta_y)+'\t'+str(theta_z)+'\t')
for value in line:
fout.write(str(value)+'\t')
fout.write(str(R[2][0])+'\t'+str(R[2][1])+'\t'+str(R[2][2])+'\t')
fout.write('\n')
fout.close()
n_clusters_ = len(np.unique(labels))
print("Number of estimated clusters:", n_clusters_)
classification = open('analysis/frame_cluster_types.txt', 'w+')
frame_counter = 0
framelist_counter = 0
for label in labels:
if frame_counter<framelist[framelist_counter]:
classification.write(str(label)+'\t')
frame_counter+=1
else:
framelist_counter+=1
classification.write('\n'+str(label)+'\t')
frame_counter=1
'''
def from4atoms(OA, OB, OC, OD, u, v, l, m, accept_no_solution=True):
u = pi - u
v = pi - v
a = OA - OC
b = OB - OD
ddiff = OC - OD + a - b # (BA) = (B-A)
x = a.norm() * l * cos(u)
y = b.norm() * m * cos(v)
rx = x
ry = y - b * ddiff
r1 = a
r2 = b
if(abs(r1[0]) < 0.000001):
if(abs(r2[0]) < 0.000001):
print "WARNING: r1 = r2 = 0.0"
# Swap rows
tmp = r1
r1 = r2
r2 = tmp
tmp = rx
ry = rx
rx = tmp
# Reduce the matrix
factor = r1[0]
r1 = r1 / factor
rx = rx / factor
factor = r2[0]
r2 = r2 - (r1 ** factor)
ry = ry - rx * factor
factor = r2[1]
r2 = r2 / factor
ry = ry / factor
factor = r1[1]
r1 = r1 - r2 ** factor
rx = rx - ry * factor
# Make solution space vectors
alpha = r1[2]
beta = r2[2]
gamma = rx
delta = ry
u = Vector(-alpha,-beta,1.0)
v = Vector(gamma, delta, 0)
# Solve quadratic equation for norm of c
acoef = (u.norm())**2
bcoef = 2*(u*v)
ccoef = (v.norm())**2 - l**2
disc = bcoef**2 - 4*acoef*ccoef
if(disc < 0.0):
# This is the sick case where we can't find _any_ solution
if not accept_no_solution:
raise ValueError, "from4atoms: no solution found (disc=%f)" % disc
else:
disc = 0
x1 = (-bcoef - sqrt(disc))/(2*acoef)
x2 = (-bcoef + sqrt(disc))/(2*acoef)
# Create the two c-vectors
c1 = u ** x1 + v
c2 = u ** x2 + v
# The two candidate E-poitns
E1 = OA + c1
E2 = OA + c2
# The two candidate d-vectors
d1 = E1 - OB
d2 = E2 - OB
# Pick the one with smallest norm difference
d1norm = d1.norm()
d2norm = d2.norm()
diff1 = abs(d1norm - m)
diff2 = abs(d2norm - m)
if(diff1 < diff2):
return E1
else:
return E2