def _download_parks(file_path, geofence_file, nest_parks=False):
parks = []
geofences = parse_geofence_file(geofence_file)
for geofence in geofences:
log.info('Downloading parks in geofence %s...', geofence['name'])
box = get_geofence_box(geofence)
parks_in_box = _query_overpass_api(box['sw'], box['ne'], nest_parks)
coords = geofence['polygon']
coords.append(coords[0])
path = matplotlib.path.Path(coords)
parks_in_geofence = []
for park in parks_in_box:
for coord in park:
coord_tuple = (coord[0], coord[1])
if path.contains_point(coord_tuple):
parks_in_geofence.append(park)
break
log.info('%d parks found in geofence %s.', len(parks_in_geofence),
geofence['name'])
parks.extend(parks_in_geofence)
output = {"date": str(datetime.now()), "parks": parks}
if len(output['parks']) > 0:
with open(file_path, 'w') as file:
json.dump(output, file, separators=(',', ':'))
log.info('%d parks downloaded to %s.', len(output['parks']), file_path)
else:
log.info('0 parks downloaded. Skipping saving to %s', file_path)
def flood_area(polygon):
path = matplotlib.path.Path(polygon)
# Calculated center.
lat, lon = [sum(y) / len(y) for y in zip(*polygon)]
starting_cell = s2sphere.CellId.from_lat_lng(s2sphere.LatLng.from_degrees(
lat, lon)).parent(16)
seen = set()
cells = set()
candidates = [starting_cell]
# Add all cells that are strictly within the polygon.
while candidates:
cell = candidates.pop()
if not path.contains_point(lat_lng(cell)):
continue
cells.add(cell)
for neighbour in cell.get_edge_neighbors():
if neighbour not in seen:
seen.add(neighbour)
candidates.append(neighbour)
# Then add all cells on the border to make sure we cover the area.
for cell in list(cells):
for neighbour in cell.get_edge_neighbors():
if neighbour not in cells:
cells.add(neighbour)
return cells
def find_containing_building(latlon):
"""Find the building that contains item"""
import matplotlib.path
for building in locations:
if building.data['type'] != 'building': continue
path = matplotlib.path.Path(building.outline)
if path.contains_point(latlon):
return building
def point_in_polygon(x: float, y: float,
polygon: typing.List[typing.List[float]]) -> bool:
"""
Checks if a point is in a given polygon.
:param x: x
:param y: y
:param polygon: The polygon to check for inclusion
:return: Whether or not the given point is in the provided polygon
"""
path = matplotlib.path.Path(vertices=numpy.array(polygon), closed=True)
return path.contains_point([x, y])
def count_by_position(regions, data, key=None):
paths = {
name: matplotlib.path.Path(
np.array(polygon),
([matplotlib.path.Path.MOVETO] +
[matplotlib.path.Path.LINETO for _ in range(len(polygon) - 2)] +
[matplotlib.path.Path.CLOSEPOLY]
)
)
for name, polygons in regions.iteritems()
for polygon in polygons
}
count = collections.defaultdict(int)
for entry in data:
for name, path in paths.iteritems():
pos = entry['position'][1], entry['position'][0]
if path.contains_point(pos):
count[name] += int(entry.get(key, 1))
return count
def is_point_within_polygon(point, polygonVertexCoords):
path = matplotlib.path.Path( polygonVertexCoords)
return path.contains_point(point[0], point[1])
def test_path():
"""Create a fake path and test it for anomalies using KDEClassifier."""
with open("points.pickle", "r") as f:
points = pickle.load(f)
# all points outside of poly are ignored
"""
poly = np.array([
[115.707, 7.9],
[116.509, 7.727],
[118.295, 10.3057],
[118.915, 13.4038],
[118.21, 13.4038],
[117.652, 10.667],
[115.707, 7.9]
])
"""
poly = np.array([[106.146, 4.0428], [114.046, 12.1145], [111.592, 14.1508],
[105.508, 4.06733], [106.146, 4.0428]])
"""
poly = np.array([
[113.531, 11.356],
[113.531, 9.2452],
[116.328, 9.2452],
[116.328, 11.356],
[113.531, 11.356]
])
"""
raw_training_data = np.array([[
p[0], p[1], p[3] * np.sin(p[2] / 10.0 * np.pi / 180) / 600,
p[3] * np.cos(p[2] / 10.0 * np.pi / 180) / 600
] for p in points])
path = matplotlib.path.Path(poly[:-1], closed=True)
training_data = raw_training_data
# crop using poly
training_data = training_data[np.apply_along_axis(
lambda x: False if path.contains_point([x[0], x[1]]) == 0 else True, 1,
raw_training_data)]
# select stationary points
# training_data = training_data[np.where(training_data[:,2]**2+training_data[:,3]**2 < 10**-2)]
kde = KDEClassifier(training_data)
minx = min(training_data[:, 0])
miny = min(training_data[:, 1])
maxx = max(training_data[:, 0])
maxy = max(training_data[:, 1])
"""
start = np.array([116.103, 7.87])
middle = np.array([117.97, 10.47])
anom1 = np.array([117.67, 11.423])
anom2 = np.array([119.221, 10.5673])
anom3 = np.array([116.402, 10.8365])
normal1 = np.array([117.786, 10.0526])
normal2 = np.array([118.224, 11.09])
anom4 = np.array([118.627, 9.99])
end = np.array([118.604, 13.234])
testing_data = np.array(create_long_path([start, normal1, anom4, normal2, end], 0.15, 0.03, 0.06))
"""
# create a fake path
start = np.array([106.085, 4.458])
end = np.array([112.875, 12.623])
normal1 = np.array([109.655, 8.7621])
normal2 = np.array([111.436, 10.8496])
normal3 = np.array([108.243, 7.10426])
anom1 = np.array([108.888, 11.248])
anom2 = np.array([108.12, 4.0956])
anom3 = np.array([106.505, 9.48133])
testing_data = np.array(create_long_path([end, start], 0.15, 0.03, 0.06))
# classify!
fixed_anom_bools = kde.classify(testing_data, fixed=True, approx=True)
# plot results
m = get_basemap(minx - 1, maxx + 1, miny - 1, maxy + 1)
xpts, ypts = m(training_data[:, 0], training_data[:, 1])
us, vs = training_data[:, 2], training_data[:, 3]
m.scatter(xpts, ypts, c="k", marker=".", alpha=0.3)
# m.quiver(xpts, ypts, us, vs, alpha=0.3)
# m.plot(poly[:,0], poly[:,1])
xs, ys = m(testing_data[:, 0], testing_data[:, 1])
# us, vs = testing_data[:,2], testing_data[:,3]
colors = np.where(fixed_anom_bools, "r", "g")
m.scatter(xs, ys, c=colors, marker="8", s=100)
# m.quiver(xs, ys, us, vs, color=colors, scale=3)
# m.hexbin(training_data[:,0], training_data[:,1])
plt.show()
#sys.exit(0)
# test_path()
points = load_points()
print "processed file"
poly = np.array([[91.0695, -6.4], [91.0695, -19.542], [120.211, -19.542],
[120.211, -6.4], [91.0695, -6.4]])
path = matplotlib.path.Path(poly, closed=True)
# points = points[-200000:,:2]
points = points[np.apply_along_axis(
lambda x: False
if path.contains_point([x[0], x[1]]) == 0 else True, 1, points)]
points = points[np.where(points[:, 2]**2 + points[:, 3]**2 == 0)][:, :2]
print points.shape
invcov = np.linalg.inv(np.cov(points.T))
# metric = sklearn.neighbors.DistanceMetric.get_metric("mahalanobis", VI=invcov)
def metric(x, y):
# why is this needed
if len(x) == 10:
return 0
return np.sqrt(np.einsum("i,ij,j", x - y, invcov, x - y))
# metric = lambda foo, bar: np.sqrt((foo-bar).dot(foo-bar))
labels = sklearn.cluster.DBSCAN(eps=0.3, metric="euclidean").fit_predict(
points[:, :2])
def test_path():
"""Create a fake path and test it for anomalies using KDEClassifier."""
with open("points.pickle", "r") as f:
points = pickle.load(f)
# all points outside of poly are ignored
"""
poly = np.array([
[115.707, 7.9],
[116.509, 7.727],
[118.295, 10.3057],
[118.915, 13.4038],
[118.21, 13.4038],
[117.652, 10.667],
[115.707, 7.9]
])
"""
poly = np.array([
[106.146, 4.0428],
[114.046, 12.1145],
[111.592, 14.1508],
[105.508, 4.06733],
[106.146, 4.0428]
])
"""
poly = np.array([
[113.531, 11.356],
[113.531, 9.2452],
[116.328, 9.2452],
[116.328, 11.356],
[113.531, 11.356]
])
"""
raw_training_data = np.array([[p[0], p[1], p[3]*np.sin(p[2]/10.0*np.pi/180)/600, p[3]*np.cos(p[2]/10.0*np.pi/180)/600] for p in points])
path = matplotlib.path.Path(poly[:-1], closed=True)
training_data = raw_training_data
# crop using poly
training_data = training_data[np.apply_along_axis(lambda x: False if path.contains_point([x[0], x[1]])==0 else True, 1, raw_training_data)]
# select stationary points
# training_data = training_data[np.where(training_data[:,2]**2+training_data[:,3]**2 < 10**-2)]
kde = KDEClassifier(training_data)
minx = min(training_data[:,0])
miny = min(training_data[:,1])
maxx = max(training_data[:,0])
maxy = max(training_data[:,1])
"""
start = np.array([116.103, 7.87])
middle = np.array([117.97, 10.47])
anom1 = np.array([117.67, 11.423])
anom2 = np.array([119.221, 10.5673])
anom3 = np.array([116.402, 10.8365])
normal1 = np.array([117.786, 10.0526])
normal2 = np.array([118.224, 11.09])
anom4 = np.array([118.627, 9.99])
end = np.array([118.604, 13.234])
testing_data = np.array(create_long_path([start, normal1, anom4, normal2, end], 0.15, 0.03, 0.06))
"""
# create a fake path
start = np.array([106.085, 4.458])
end = np.array([112.875, 12.623])
normal1 = np.array([109.655, 8.7621])
normal2 = np.array([111.436, 10.8496])
normal3 = np.array([108.243, 7.10426])
anom1 = np.array([108.888, 11.248])
anom2 = np.array([108.12, 4.0956])
anom3 = np.array([106.505, 9.48133])
testing_data = np.array(create_long_path([end, start], 0.15, 0.03, 0.06))
# classify!
fixed_anom_bools = kde.classify(testing_data, fixed=True, approx=True)
# plot results
m = get_basemap(minx-1, maxx+1, miny-1, maxy+1)
xpts, ypts = m(training_data[:,0], training_data[:,1])
us, vs = training_data[:,2], training_data[:,3]
m.scatter(xpts, ypts, c="k", marker=".", alpha=0.3)
# m.quiver(xpts, ypts, us, vs, alpha=0.3)
# m.plot(poly[:,0], poly[:,1])
xs, ys = m(testing_data[:,0], testing_data[:,1])
# us, vs = testing_data[:,2], testing_data[:,3]
colors = np.where(fixed_anom_bools, "r", "g")
m.scatter(xs, ys, c=colors, marker="8", s=100)
# m.quiver(xs, ys, us, vs, color=colors, scale=3)
# m.hexbin(training_data[:,0], training_data[:,1])
plt.show()
points = load_points()
print "processed file"
poly = np.array([
[91.0695, -6.4],
[91.0695, -19.542],
[120.211, -19.542],
[120.211, -6.4],
[91.0695, -6.4]
])
path = matplotlib.path.Path(poly, closed=True)
# points = points[-200000:,:2]
points = points[np.apply_along_axis(lambda x: False if path.contains_point([x[0], x[1]])==0 else True, 1, points)]
points = points[np.where(points[:,2]**2+points[:,3]**2 == 0)][:,:2]
print points.shape
invcov = np.linalg.inv(np.cov(points.T))
# metric = sklearn.neighbors.DistanceMetric.get_metric("mahalanobis", VI=invcov)
def metric(x, y):
# why is this needed
if len(x) == 10:
return 0
return np.sqrt(np.einsum("i,ij,j", x-y, invcov, x-y))
# metric = lambda foo, bar: np.sqrt((foo-bar).dot(foo-bar))
labels = sklearn.cluster.DBSCAN(eps=0.3, metric="euclidean").fit_predict(points[:,:2])
print labels
unique_labels = list(set(labels))