def get_graph_from_geometry(geometry):
# let's transform the geometry into lists of channel names and coordinates
chans, coords = zip(*[(ch, xy) for ch, xy in geometry.iteritems()])
# we'll perform the triangulation and extract the
try:
tri = spatial.Delaunay(coords)
except QhullError:
# oh no! we probably have a linear geometry.
chans, coords = list(chans), list(coords)
x, y = zip(*coords)
# let's add a dummy channel and try again
coords.append((max(x) + 1, max(y) + 1))
tri = spatial.Delaunay(coords)
# then build the list of edges from the triangulation
indices, indptr = tri.vertex_neighbor_vertices
edges = []
for k in range(indices.shape[0] - 1):
for j in indptr[indices[k]:indices[k + 1]]:
try:
edges.append((chans[k], chans[j]))
except IndexError:
# let's ignore anything connected to the dummy channel
pass
return edges
def get_graph_from_geometry(geometry):
# let's transform the geometry into lists of channel names and coordinates
chans, coords = zip(*[(ch, xy) for ch, xy in geometry.iteritems()])
# we'll perform the triangulation and extract the
try:
tri = spatial.Delaunay(coords)
except:
x, y = zip(*coords)
coords = list(coords)
coords.append((max(x) + 1, max(y) + 1))
tri = spatial.Delaunay(coords)
# then build the list of edges from the triangulation
indices, indptr = tri.vertex_neighbor_vertices
print indices, indptr
edges = []
for k in range(indices.shape[0] - 1):
for j in indptr[indices[k]:indices[k + 1]]:
try:
edges.append((chans[k], chans[j]))
except IndexError:
# ignore dummy site
pass
return edges
def is_in_hull(
points: ArrayLike, hull: Union[ss.Delaunay, ss.ConvexHull,
ArrayLike]) -> np.ndarray:
"""
test if points in `points` are in `hull`
Parameters
----------
points: array_like,
an `NxK` coordinates of `N` points in `K` dimensions
hull: Delaunay object or ConvexHull object, or array_like,
the objects which defines the hull, essentially an `MxK` array of the coordinates of `M` points in `K`dimensions
Returns
-------
ndarray of bool
"""
if isinstance(hull, ss.Delaunay):
_h = hull
elif isinstance(hull, ss.ConvexHull):
_h = ss.Delaunay(hull.points)
else:
_h = ss.Delaunay(hull)
return _h.find_simplex(points) >= 0
def estimate(self, src, dst):
"""Set the control points with which to perform the piecewise mapping.
Number of source and destination coordinates must match.
Parameters
----------
src : (N, 2) array
Source coordinates.
dst : (N, 2) array
Destination coordinates.
"""
# forward piecewise affine
# triangulate input positions into mesh
self._tesselation = spatial.Delaunay(src)
# find affine mapping from source positions to destination
self.affines = []
for tri in self._tesselation.vertices:
affine = AffineTransform()
affine.estimate(src[tri, :], dst[tri, :])
self.affines.append(affine)
# inverse piecewise affine
# triangulate input positions into mesh
self._inverse_tesselation = spatial.Delaunay(dst)
# find affine mapping from source positions to destination
self.inverse_affines = []
for tri in self._inverse_tesselation.vertices:
affine = AffineTransform()
affine.estimate(dst[tri, :], src[tri, :])
self.inverse_affines.append(affine)
def match_candidates_by_graph(
images_ref: List[str],
images_cand: List[str],
exifs: Dict[str, Any],
reference: geo.TopocentricConverter,
rounds: int,
):
"""Find by triangulating the GPS points on X/Y axises"""
if len(images_cand) == 0 or rounds < 1:
return set()
images_cand_set = set(images_cand)
images_ref_set = set(images_ref)
images = list(images_cand_set | images_ref_set)
representative_points = get_representative_points(images, exifs, reference)
points = np.zeros((len(images), 2))
for i, point in enumerate(representative_points.values()):
points[i] = point[0:2]
def produce_edges(triangles):
for triangle in triangles:
for vertex1, vertex2 in combinations(triangle, 2):
image1, image2 = images[vertex1], images[vertex2]
if image1 == image2:
continue
pair_way1 = image1 in images_cand_set and image2 in images_ref_set
pair_way2 = image2 in images_cand_set and image1 in images_ref_set
if pair_way1 or pair_way2:
yield sorted_pair(image1, image2), (vertex1, vertex2)
pairs = set()
# first round compute scale based on edges (and push delaunay edges)
edge_distances = []
triangles = spatial.Delaunay(points).simplices
for (image1, image2), (vertex1, vertex2) in produce_edges(triangles):
pairs.add((image1, image2))
edge_distances.append(norm_2d(points[vertex1] - points[vertex2]))
scale = np.median(edge_distances)
# further rounds produces edges from jittered version of the original points
# in order to get 'alternative' delaunay triangulations : a perfect square
# will only produce one diagonal edge, so by jittering it, we get more
# chances of getting such diagonal edges and having more diversity
for _ in range(rounds):
points_current = copy.copy(points) + np.random.rand(*points.shape) * scale
triangles = spatial.Delaunay(points_current).simplices
for (image1, image2), _ in produce_edges(triangles):
pairs.add((image1, image2))
return pairs
def evaluate(self, points, values, rpoints, method="linear"):
if len(np.array(points).shape) == 1:
self._dim = 1
points = points[:, np.newaxis]
elif len(points.shape) == 2:
self._dim = points.shape[1]
else:
raise ValueError("Points must be ndarray of shape (npoints,dim)")
methods = ["nearest", "linear", "cubic"]
if not method in methods:
raise ValueError("Supported methods: {}".format(
", ".join(methods)))
if method == "cubic" and self._dim > 2:
raise ValueError(
"Cubic interpolator only supported for 1D and 2D data")
nanmask = ~np.isnan(values)
if not np.array_equal(self._points, points) \
or self._method != method \
or not np.array_equal(self._nanmask, nanmask):
self._kdtree = spatial.cKDTree(points)
if self._dim == 1:
self._triangulation = None
elif method == "nearest":
self._triangulation = None
elif method == "linear":
self._triangulation = spatial.Delaunay(points)
elif method == "cubic":
self._triangulation = spatial.Delaunay(points[nanmask])
self._points = points
self._method = method
self._nanmask = nanmask
if method == "nearest":
ivalues = self._nearest_interpolator(rpoints, values)
elif method == "linear":
lval = self._linear_interpolator(rpoints, values)
mask = np.isnan(lval)
if np.any(mask):
lval[mask] = self._nearest_interpolator(
[x[mask] for x in rpoints], values)
ivalues = lval
elif method == "cubic":
cval = self._cubic_interpolator(rpoints, values)
nval = self._nearest_interpolator(rpoints, values)
cval[np.isnan(nval)] = np.nan
ivalues = cval
return np.squeeze(ivalues)
def getBlendedImage(self, alpha):
trianglePoints = spatial.Delaunay(self.endPoints).vertices
sourceTarget = np.zeros(self.startImage.shape, dtype=np.uint8)
endTarget = np.zeros(self.endImage.shape, np.uint8)
for vertices in trianglePoints:
# ALL COORDINATES ARE IN X,Y
sourceP1 = (self.startPoints[vertices[0]][0],
self.startPoints[vertices[0]][1])
sourceP2 = (self.startPoints[vertices[1]][0],
self.startPoints[vertices[1]][1])
sourceP3 = (self.startPoints[vertices[2]][0],
self.startPoints[vertices[2]][1])
endP1 = (self.endPoints[vertices[0]][0],
self.endPoints[vertices[0]][1])
endP2 = (self.endPoints[vertices[1]][0],
self.endPoints[vertices[1]][1])
endP3 = (self.endPoints[vertices[2]][0],
self.endPoints[vertices[2]][1])
destP1 = ((1 - alpha) * sourceP1[0] + alpha * endP1[0],
(1 - alpha) * sourceP1[1] + alpha * endP1[1])
destP2 = ((1 - alpha) * sourceP2[0] + alpha * endP2[0],
(1 - alpha) * sourceP2[1] + alpha * endP2[1])
destP3 = ((1 - alpha) * sourceP3[0] + alpha * endP3[0],
(1 - alpha) * sourceP3[1] + alpha * endP3[1])
sourceAffine = Affine(np.array([sourceP1, sourceP2, sourceP3]),
np.array([destP1, destP2, destP3]))
sourceAffine.transform(self.startImage, sourceTarget)
endAffine = Affine(np.array([endP1, endP2, endP3]),
np.array([destP1, destP2, destP3]))
endAffine.transform(self.endImage, endTarget)
return np.array((1 - alpha) * sourceTarget + alpha * endTarget,
dtype=np.uint8)
def alpha_shape_auto(xys, step=1, verbose=False):
triangulation = spat.Delaunay(xys)
triangles = xys[triangulation.simplices]
a_pts = triangles[:, 0, :]
b_pts = triangles[:, 1, :]
c_pts = triangles[:, 2, :]
radii = r_circumcircle_triangle(a_pts, b_pts, c_pts)
radii[np.isnan(radii)] = 0 # "Line" triangles to be kept for sure
del triangles, a_pts, b_pts, c_pts
radii_sorted_i = radii.argsort()
triangles = triangulation.simplices[radii_sorted_i][::-1]
radii = radii[radii_sorted_i][::-1]
geoms_prev = alpha_geoms((1 / radii.max()) - 1e-10, triangles, radii, xys)
xys_bb = np.array([*xys.min(axis=0), *xys.max(axis=0)])
if verbose:
print('Step set to %i' % step)
for i in range(0, len(radii), step):
radi = radii[i]
alpha = (1 / radi) - 1e-100
if verbose:
print('%.2f%% | Trying a = %f'\
%((i+1)/radii.shape[0], alpha))
geoms = alpha_geoms(alpha, triangles, radii, xys)
if (geoms.shape[0] != 1) or not (np.all(xys_bb == geoms.total_bounds)):
break
else:
geoms_prev = geoms
return geoms_prev[0] # Return a shapely polygon
def generate_throats(self):
r"""
Generate the throats (connections, numbering and types)
"""
self._logger.info("generate_throats: Define connections between pores")
Np = self._net.get_num_pores()
pts = self._net.pore_properties['coords']
#Generate 6 dummy domains to pad onto each face of real domain
#This prevents surface pores from making long range connections to each other
Lx = self.domain_size[0]
Ly = self.domain_size[1]
Lz = self.domain_size[2]
f = 0.1; #Scale factor for size of dummy domains
ptsX0 = np.hstack((-np.random.rand(Np*f,1)*Lx*f, np.random.rand(Np*f,1)*Ly, np.random.rand(Np*f,1)*Lz))
ptsY0 = np.hstack((np.random.rand(Np*f,1)*Lx, -np.random.rand(Np*f,1)*Ly*f, np.random.rand(Np*f,1)*Lz))
ptsZ0 = np.hstack((np.random.rand(Np*f,1)*Lx, np.random.rand(Np*f,1)*Ly, -np.random.rand(Np*f,1)*Lz*f))
ptsXX = np.hstack((np.random.rand(Np*f,1)*Lx*f+Lx, np.random.rand(Np*f,1)*Ly, np.random.rand(Np*f,1)*Lz))
ptsYY = np.hstack((np.random.rand(Np*f,1)*Lx, np.random.rand(Np*f,1)*Ly*f+Ly, np.random.rand(Np*f,1)*Lz))
ptsZZ = np.hstack((np.random.rand(Np*f,1)*Lx, np.random.rand(Np*f,1)*Ly, np.random.rand(Np*f,1)*Lz*f+Lz))
#Add dummy domains to real domain
pts = np.concatenate([pts,ptsX0,ptsXX,ptsY0,ptsYY,ptsZ0,ptsZZ])
#Perform tessellation
Tri = sptl.Delaunay(pts)
adjmat = sprs.lil_matrix((Np,Np),dtype=int)
for i in np.arange(0,np.shape(Tri.simplices)[0]):
#Keep only simplices that are fully in real domain
adjmat[Tri.simplices[i][Tri.simplices[i]<Np],Tri.simplices[i][Tri.simplices[i]<Np]] = 1
#Remove duplicate (lower triangle) and self connections (diagonal)
#and convert to coo
adjmat = sprs.triu(adjmat,k=1,format="coo")
self._net.throat_properties['connections'] = np.vstack((adjmat.row, adjmat.col)).T
self._net.throat_properties['type'] = np.zeros_like(adjmat.row)
self._net.throat_properties['numbering'] = np.arange(0,np.size(adjmat.row))
self._logger.debug("generate_throats: End of method")
def min_dist(point_list):
"""
Find the minimum distance between two points in a point list using
Delaunay triangulation.
"""
# Find the Delaunay triangulation
mesh = spatial.Delaunay(point_list)
# Get the edges of the triangulation
edges = np.vstack((mesh.simplices[:, :2], mesh.simplices[:, -2:]))
# The x- and y- coordinates of the edges
x = mesh.points[edges[:, 0]]
y = mesh.points[edges[:, 1]]
# Calculate the lengths of the Delaunay edges, which are candidates for minimum distance
dists = np.sqrt(np.sum((x - y)**2, axis=1))
idx = np.argmin(dists)
idx1, idx2 = edges[idx]
p1, p2 = point_list[edges[idx]]
print('The closest points are:')
print('Index: {}, coordinates: {}'.format(idx1, tuple(p1)))
print('Index: {}, coordinates: {}'.format(idx2, tuple(p2)))
return np.min(dists)
def _compute_hull(self, qhull_options):
"""Compute properties of the convex hull and set up the outer hull.
"""
self.inside_vertices = self.in_hull(self.vor.vertices)
self.hull_points = self._flatten_simplices(self.hull.convex_hull)
self.hull_center = np.mean(self.hull.points[self.hull_points], axis=0)
# enlarge hull:
# estimate area needed for a poisson process with same mean distance:
new_area = 4.0 * self.mean_nearest_distance**2 * self.vor.npoints
fac = np.sqrt(new_area / self.hull_area())
self.outer_hull_points = np.zeros(
(len(self.hull_points), self.vor.ndim))
for k, point in enumerate(self.hull.points[self.hull_points]):
point -= self.hull_center
point *= fac
point += self.hull_center
self.outer_hull_points[k] = point
# compute outer hull:
self.outer_hull = sp.Delaunay(self.outer_hull_points,
furthest_site=False,
incremental=False,
qhull_options=qhull_options)
self.outer_min_bound = np.min(self.outer_hull_points, axis=0)
self.outer_max_bound = np.max(self.outer_hull_points, axis=0)
def _generate_mesh(self, env):
# generate nodes, first node is the goal
points = env._generate_points(self._sample_num - 2)
points = np.concatenate(([[env._end.x, env._end.y], [env._start.x, env._start.y]], points))
while True:
dist_list = np.array([Point(xy[0], xy[1]).distance(env.obstacles) for xy in points])
collision_mask = dist_list < 1e-5
collision_count = np.sum(collision_mask)
if collision_count == 0:
break
# resample
points[collision_mask] = env._generate_points(collision_count)
self._samples = MultiPoint(points)
# generate triangles
tesselation = sps.Delaunay(points)
triangles = tesselation.simplices.copy()
edges = set()
for tri in triangles:
sortnodes = np.sort(tri)
edges.add((sortnodes[0], sortnodes[1]))
edges.add((sortnodes[1], sortnodes[2]))
edges.add((sortnodes[0], sortnodes[2]))
line_list = []
obstacle_union = unary_union(env.obstacles)
for n1, n2 in edges:
line = LineString([self._samples[n1], self._samples[n2]])
if line.intersection(obstacle_union).is_empty:
line_list.append((n1, n2))
self._connections = line_list
def rectangle(cls, parent, height, width, pos_com=(0.,0.)):
'''
Generate a rectangle of given height, width and center of mass.
Parameters
----------
parent : :class:`~nngt.SpatialGraph` or subclass
The parent container.
height : float
Height of the rectangle.
width : float
Width of the rectangle.
pos_com : tuple of floats, optional (default: (0., 0.))
Position of the rectangle's center of mass
Returns
-------
shape : :class:`~nngt.Shape`
Rectangle shape.
'''
shape = cls(parent)
half_diag = np.sqrt(np.square(height/2.) + np.square(width/2.)) / 2.
pos_com = np.array(pos_com)
points = [ pos_com + [half_diag,half_diag],
pos_com + [half_diag,-half_diag],
pos_com - [half_diag,half_diag],
pos_com - [half_diag,-half_diag] ]
shape._convex_hull = sptl.Delaunay(points)
shape._com = pos_com
shape._area = height * width
return shape
def build_tri(self):
"""
Create a matplotlib triangulation object from the grid
Used primarily to contour the data (but also for interpolation)
"""
if '_tri' not in self.__dict__:
maxfaces = self.nfaces.max()
if maxfaces == 3:
self._tri = tri.Triangulation(self.xp, self.yp, self.cells)
else:
# Need to compute a delaunay triangulation for mixed meshes
pts = np.vstack([self.xp, self.yp]).T
D = spatial.Delaunay(pts)
self._tri = tri.Triangulation(self.xp, self.yp, D.simplices)
# Compute a mask by removing triangles outside of the polygon
xy = D.points
cells = D.simplices
xv,yv = circumcenter(xy[cells[:,0],0],\
xy[cells[:,0],1],\
xy[cells[:,1],0],\
xy[cells[:,1],1],\
xy[cells[:,2],0],\
xy[cells[:,2],1])
mask = self.find_cell(xv, yv)
self._tri.set_mask(mask == -1)
def _interp_functor(interpolator, xi, points, *points_props):
points = np.array(points)
points_props = tuple(map(np.array, points_props))
xi = tuple(map(np.array, xi))
delaunay = (spatial.Delaunay(points) if points.shape[1] >= 2
else Delaunay1D(points[:, 0]))
simplex_map = delaunay.find_simplex(np.stack(xi, axis=-1))
simplex_outer_loc = np.unique(delaunay.convex_hull)
simplex_outer_base = points[simplex_outer_loc]
simplex_outer_args = tuple(arg[simplex_outer_loc] for arg in points_props)
interp = (interpolator(xi, simplex_outer_base, *simplex_outer_args)
if len(simplex_outer_loc) == points.shape[0] - 1
else np.full(xi[0].shape, np.nan))
for i, s in enumerate(delaunay.simplices):
mask = simplex_map == i
masked_xi = tuple(dim[mask] for dim in xi)
simplex_base = points[s]
simplex_args = tuple(arg[s] for arg in points_props)
interp[mask] = interpolator(masked_xi, simplex_base, *simplex_args)
return interp
def triangleRemoval(triangulation, heights, K):
"""
removes triangulation from the delaunay triangulation
:param triangulation: delaunay triangulation object
:param heights: 1d array of the heights of the triangle points
:param K: number of triangles to keep
:return: simplified triangulation using triangle removal method
"""
deleted_points = []
while triangulation.simplices.shape[0] > K:
area = computeTriArea(triangulation)
tri_to_delete = triangulation.simplices[np.argmin(area)]
points_to_delete = tri_to_delete # indices of the points of the triangle we deleting
deleted_points.append(
np.hstack((np.take(triangulation.points, tri_to_delete,
axis=0), np.take(heights, tri_to_delete,
axis=0))))
# deleting the chosen points and their heights
temp_points = np.delete(triangulation.points, [*points_to_delete],
axis=0)
heights = np.delete(heights, [*points_to_delete], axis=0)
triangulation = spat.Delaunay(temp_points)
print("number of triangles is ",
triangulation.simplices.shape[0]) # progress indication
return triangulation, heights, np.vstack(deleted_points)
def vertexRemoval(triangulation, heights, K):
"""
detects and removes a vertex from a given set of points. returns an updated delaunay triangulation
:param triangulation: delaunay triangulation object
:param heights: 1d array of the heights of the triangle points
:param K: number of triangles to keep
:return: simplified triangulation using vertex removal method
"""
deleted_points = []
while triangulation.simplices.shape[0] > K:
area = computeTriArea(triangulation)
tri_to_delete = triangulation.simplices[np.argmin(area)]
points_to_delete = tri_to_delete[
0] # indices of the points of the triangle we are deleting
deleted_points.append(
np.hstack((triangulation.points[points_to_delete],
heights[points_to_delete])))
# deleting the chosen points and their heights
temp_points = np.delete(triangulation.points, points_to_delete, axis=0)
heights = np.delete(heights, points_to_delete, axis=0)
triangulation = spat.Delaunay(temp_points)
print("number of triangles is ",
triangulation.simplices.shape[0]) # progress indication
return triangulation, heights, np.vstack(deleted_points)
def _find_delaunay(df, parameters=None, call_num=None):
method_key = get_method_key('neighbours')
cutoff = get_param_val(parameters[method_key]['cutoff'])
points = df[['x', 'y']].values
particle_ids = df[['particle']].values.flatten()
tess = sp.Delaunay(points)
list_indices, point_indices = tess.vertex_neighbor_vertices
# neighbour_ids = [particle_ids[point_indices[a:b].tolist()] for a, b in zip(list_indices[:-1], list_indices[1:])]
neighbour_ids = [
point_indices[a:b].tolist()
for a, b in zip(list_indices[:-1], list_indices[1:])
]
dist = sp.distance.squareform(sp.distance.pdist(points))
neighbour_dists = [(dist[i, row] < cutoff).tolist()
for i, row in enumerate(neighbour_ids)]
indices = []
for index, row in enumerate(neighbour_ids):
indices.append([
particle_ids[neighbour_ids[index][j]]
for j, dummy in enumerate(row) if neighbour_dists[index][j]
])
df.loc[:, ['neighbours']] = indices
return df
def refine_by_delaunay(self, triangles):
"""refines given elements by the delaunay refinement method"""
p = self.node_list
t = self.face_list
for tr in triangles:
pmid = (p[t[tr][0]] + p[t[tr][1]] + p[t[tr][2]]) / 3
p = np.append(p, [pmid], axis=0)
self.edge_length /= 2
dp = self.edge_length
for i in range(1, int(ceil(1 / dp))):
p = np.append(p, [[i * dp, 0]], axis=0)
for i in range(int(ceil(.5 / dp))):
p = np.append(p, [[1, i * dp]], axis=0)
for i in range(int(ceil(.5 / dp))):
p = np.append(p, [[1 - i * dp, .5]], axis=0)
for i in range(int(ceil(.5 / dp))):
p = np.append(p, [[.5, .5 + i * dp]], axis=0)
for i in range(int(ceil(.5 / dp))):
p = np.append(p, [[.5 - i * dp, 1]], axis=0)
for i in range(int(ceil(1 / dp))):
p = np.append(p, [[0, 1 - i * dp]], axis=0)
p = unique2d(p)
t = spspatial.Delaunay(p).vertices
# remove triangles where centroid outside boundary (good enough...)
pmid = p[t].sum(1) / 3
t = t[self.fd(pmid) < 1e-15]
t = [Face(t[i]) for i in range(len(t))]
self.node_list = p
self.face_list = t
def delaunay(points, shape=[1, 1, 1]):
r"""
Generate a network based on Delaunay triangulation of random points
Parameters
----------
points : array_like or int
Can either be an N-by-3 array of point coordinates which will be used,
or a scalar value indicating the number of points to generate
shape : array_like
Indicates the size and shape of the domain
Returns
-------
network : dict
A dictionary containing 'vert.coords' and 'edge.conns'
tri : Delaunay tessellation object
The Delaunay tessellation object produced by ``scipy.spatial.Delaunay``
"""
points = tools.parse_points(points=points, shape=shape)
mask = ~np.all(points == 0, axis=0)
tri = sptl.Delaunay(points=points[:, mask])
coo = tri_to_am(tri)
d = {}
d['vert.coords'] = points
d['edge.conns'] = np.vstack((coo.row, coo.col)).T
return d, tri
def make(self, key):
density = 110000 # per cubic mm
xyz = np.stack((design.Geometry.EPixel() & key).fetch('e_loc'))
margin = 75
bounds_min = xyz.min(axis=0) - margin
bounds_max = xyz.max(axis=0) + margin
volume = (bounds_max - bounds_min).prod() * 1e-9
npoints = int(volume * density + 0.5)
# generate random points that aren't too close
min_distance = 10.0 # cells aren't not allowed any closer
points = np.empty((npoints, 3), dtype='float32')
replace = np.r_[:npoints]
while replace.size:
points[replace, :] = np.random.rand(replace.size, 3) * (bounds_max - bounds_min) + bounds_min
replace = spatial.cKDTree(points).query_pairs(min_distance, output_type='ndarray')[:, 0]
# eliminate points that are too distant
inner = (spatial.Delaunay(xyz).find_simplex(points)) != -1
d, _ = spatial.cKDTree(points[inner, :]).query(points[~inner, :], distance_upper_bound=margin)
points = np.vstack((points[inner, :], points[~inner, :][d < margin, :]))
self.insert1(dict(
key, margin=margin,
density=density,
npoints=points.shape[0], min_distance=min_distance,
points=points,
volume=spatial.ConvexHull(points).volume * 1e-9,
inner_count=inner.sum()))
def DrawVoronoiTesselation(hdf5_file, frame=0):
"""
wanted structure: np.array([[3, 0], [4.2, 1], [0, 8], ...])
:return:
"""
x_gfp, y_gfp, x_rfp, y_rfp = GetXandYcoordinatesPerFrame(hdf5_file=hdf5_file, frame=frame)
#x_gfp, y_gfp, x_rfp, y_rfp = ConvertXandYcoordinates(x_gfp, y_gfp, x_rfp, y_rfp)
# Restructure so that you get a matrix; merge channels as the cell identity is not important in this case:
points = []
for x, y in zip(x_gfp + x_rfp, y_gfp + y_rfp):
points.append([x, y])
points = np.array(points)
vor = sp.Voronoi(points)
tri = sp.Delaunay(points)
fig = sp.voronoi_plot_2d(vor)
fig = sp.delaunay_plot_2d(tri)
plt.grid(which='major')
plt.show()
plt.close()
return vor, tri
def mkPlns(plnOr, org=[0, 0, 0], bxSdLen=1, scl=1, res=100, transp=0.5):
edge = True
smooth = True
shader = 'shaded'
glOpt = 'translucent'
plnClr = [1, 0, 1, transp]
# plnVert=np.array([list(itTl.product([0,1],repeat=3))])[0]
if np.array_equal(plnOr, np.array([1, 1, 1])):
plnVert = np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]])
# plnFc=
elif np.array_equal(plnOr, np.array([1, 1, 0])):
plnVert = np.array([[1, 0, 0], [1, 1, 0], [0, 1, 1], [0, 0, 1]])
elif np.array_equal(plnOr, np.array([1, 0, 0])):
plnVert = np.array([[1, 0, 0], [1, 1, 1], [1, 1, 0], [1, 0, 1]])
plnFc = spatial.Delaunay(plnVert)
plnDat = gl.MeshData(vertexes=plnVert, faces=plnFc.convex_hull)
plnObj = gl.GLSurfacePlotItem(plnVert,
drawEdges=edge,
smooth=smooth,
color=(plnClr[0], plnClr[1], plnClr[2],
plnClr[3]),
shader=shader,
glOptions=glOpt)
plnObj.translate(org[0] + bxSdLen / 2, org[0] + bxSdLen / 2,
org[0] + bxSdLen / 2)
return plnObj
def voronoiVessels(nPoints=10):
# Generate a bunch of points that pass the box extents slightly
xyz = random.uniform(-1.2, 1.2, size=(nPoints, 3))
# Triangulate them
dt = spatial.Delaunay(xyz)
# Set of already traced edges (2x speedup)
tracedEdges = set()
solid = np.zeros((100, 100, 100), dtype=np.uint8)
# Go throrough the tetrahedra
for n, tet in enumerate(dt.vertices):
print(n, dt.vertices.shape)
for eTuple in itertools.permutations(tet, 2):
# Sort the edge indices to make matches
eList = list(eTuple)
eList.sort()
eRef = "%i-%i" % tuple(eList)
# If we already traced this edge, skip it
if eRef in tracedEdges:
continue
tracedEdges.add(eRef)
# Dereference the points of interest
idx1, idx2 = eList
pt1 = xyz[idx1, :]
pt2 = xyz[idx2, :]
# Trace the line onto our list
solid |= doLineSegment(pt1, pt2)
return solid
def __init__(self, startImage, startPoints, endImage, endPoints):
# Validate startImage input.
if not isinstance(startImage, np.ndarray):
raise TypeError("startImage must be a numpy array.")
# Validate startPoints input.
if not isinstance(startPoints, np.ndarray):
raise TypeError("startPoints must be a numpy array.")
# Validate endImage input.
if not isinstance(endImage, np.ndarray):
raise TypeError("endImage must be a numpy array.")
# Validate endPoints input.
if not isinstance(endPoints, np.ndarray):
raise TypeError("endPoints must be a numpy array.")
# Define triangulation.
simplices = spatial.Delaunay(startPoints).simplices
# Initialize blender data.
self.startImage = startImage
self.startPoints = startPoints
self.endImage = endImage
self.endPoints = endPoints
self.simplices = simplices
def checkTriangles(self):
if self.chkTriangles.isChecked():
if self.startPoints is not None and self.startPoints.shape[0] == self.endPoints.shape[0] and self.startPoints.shape[0] > 2:
# Initialize triangle color and simplices.
if self.hasPresetPoints and self.hasCustomPoints: pen = QPen(QtCore.Qt.cyan)
elif self.hasPresetPoints: pen = QPen(QtCore.Qt.red)
else: pen = QPen(QtCore.Qt.blue)
self.simplices = spatial.Delaunay(self.startPoints).simplices
for triangle in self.simplices:
# Draw triangles in start.
sTri = QPolygonF([QPointF(self.startPoints[triangle[0]][0], self.startPoints[triangle[0]][1]),
QPointF(self.startPoints[triangle[1]][0], self.startPoints[triangle[1]][1]),
QPointF(self.startPoints[triangle[2]][0], self.startPoints[triangle[2]][1])])
self.startScene.addPolygon(sTri, pen)
# Draw triangles in end.
dTri = QPolygonF([QPointF(self.endPoints[triangle[0]][0], self.endPoints[triangle[0]][1]),
QPointF(self.endPoints[triangle[1]][0], self.endPoints[triangle[1]][1]),
QPointF(self.endPoints[triangle[2]][0], self.endPoints[triangle[2]][1])])
self.endScene.addPolygon(dTri, pen)
else:
# Remove triangles from start.
for item in self.startScene.items():
if type(item) is QGraphicsPolygonItem:
self.startScene.removeItem(item)
# Remove triangles from end.
for item in self.endScene.items():
if type(item) is QGraphicsPolygonItem:
self.endScene.removeItem(item)
def in_tetrahedron(self, t, p, true_interior=False):
"""Checks if the point P, pointed to by vector p, is inside(including the surface) the tetrahedron
If 'p' is not guaranteed a true tetrahedron, use interior().
:param t: ndarray
The four points of a tetrahedron
:param p: ndarray
The point to be tested for inclusion in the tetrahedron.
:param true_interior: bool
Activate to exclude the surface of the tetrahedron from the search.
:return: Bool
True if q is inside or on the surface of the tetrahedron.
"""
# If the surface is to be excluded, return False if p is on the surface.
if true_interior and (self.in_triangle(np.delete(t, 0, 0), p)
or self.in_triangle(np.delete(t, 1, 0), p)
or self.in_triangle(np.delete(t, 2, 0), p)
or self.in_triangle(np.delete(t, 3, 0), p)):
return False
# Check if 'p' is in the tetrahedron.
hull = spatial.Delaunay(
t) # Generate a convexHull representation of the points
return hull.find_simplex(p) >= 0 # return True if 'p' is a vertex.
def avgIm3(im1, im2, im3, pts1, pts2, pts3):
'''
compute the morph result of three images
'''
imout = np.zeros(im1.shape, dtype=np.int)
pts_avg = 1 / 3 * pts1 + 1 / 3 * pts2 + 1 / 3 * pts3
tri = spatial.Delaunay(pts_avg)
tri_pts = pts_avg[tri.simplices]
tri_pts1 = pts1[tri.simplices]
tri_pts2 = pts2[tri.simplices]
tri_pts3 = pts3[tri.simplices]
for i in range(len(tri_pts)):
transfrom1 = solveAffineTransform(tri_pts[i], tri_pts1[i])
transfrom2 = solveAffineTransform(tri_pts[i], tri_pts2[i])
transfrom3 = solveAffineTransform(tri_pts[i], tri_pts3[i])
insiders = wholePtsInTriangle(tri_pts[i])
for insider in insiders:
inv1 = np.dot(transfrom1, insider)
inv2 = np.dot(transfrom2, insider)
inv3 = np.dot(transfrom3, insider)
imout[insider[0]][insider[1]] = \
(1/3* computeBilinear(inv1, im1) + \
1/3 * computeBilinear(inv2, im2) + \
1/3 * computeBilinear(inv3, im3)).astype(int)
return imout
def mkOctHdr(org=[0,0,0],bxSdLen=1,scl=1,transp=0.5): clr=[0, 1, 1, transp] edge=True smooth=True shader='shaded' glOpt='translucent' trOcth=np.array([list(itTl.permutations([0.25,0.5,0.75,1],4))])[0] norm4du=np.array([np.sqrt(2)/2, -np.sqrt(2)/2, 0, 0]) norm4dv=np.array([np.sqrt(6)/6, np.sqrt(6)/6, -np.sqrt(2/3), 0]) norm4dw=np.array([np.sqrt(12)/12, np.sqrt(12)/12, np.sqrt(12)/12, -np.sqrt(3)/2]) ocThVert=np.array([scl*(np.sum(np.multiply(trOcth,np.tile(norm4du, (trOcth.shape[0],1))),axis=1)+0.5+org[0]), scl*(np.sum(np.multiply(trOcth,np.tile(norm4dv, (trOcth.shape[0],1))),axis=1)+0.5+org[1]), scl*(np.sum(np.multiply(trOcth,np.tile(norm4dw, (trOcth.shape[0],1))),axis=1)+0.5+org[2])]).T ocThHull=spatial.Delaunay(ocThVert) ocThObj=gl.MeshData(vertexes=ocThVert,faces=ocThHull.convex_hull) ocGrObj=gl.GLMeshItem(meshdata=ocThObj, drawEdges=edge, smooth=smooth, color=(clr[0], clr[1], clr[2], clr[3]), shader=shader, glOptions=glOpt) ocGrObj.translate(-((bxSdLen/2)+org[0]), -((bxSdLen/2)+org[1]), -((bxSdLen/2)+org[2])) ocGrObj.rotate(55,1,0,0) ocGrObj.rotate(45,0,0,1) ocGrObj.translate(((bxSdLen/2)+org[0]), ((bxSdLen/2)+org[1]), ((bxSdLen/2)+org[2])) return ocGrObj
def obtain_boundary(trj):
t0 = trj.index.get_level_values("time").unique().values[0]
Delaunay = spa.Delaunay(trj.loc[idx[t0, :], :].values)
boundary = list(np.unique(Delaunay.convex_hull.flatten()))
boundary = list(
trj.loc[idx[t0, :], :].index.get_level_values("id").values[boundary])
return boundary
