def spatial_selection(rs, re, ds, de):
if not same_direction(rs, re, ds, de, 30):
return False
lambda_r = haversine_distance(rs, re) / 2 * 1.3
middle_p = Point((rs.lat + re.lat) / 2, (rs.lon + re.lon) / 2, None)
if lambda_r > haversine_distance(
middle_p, ds) and lambda_r > haversine_distance(middle_p, de):
return False
return True
def staypoints_geolife(traj):
time_thresh = 30 * 60
dist_thresh = 250
staypoints = list()
i, i_max = 0, len(traj)
while i < i_max:
j = i + 1
token = 0
while j < i_max:
dist = haversine_distance(traj[i], traj[j])
if dist > dist_thresh:
delta_time = traj[j].datetime - traj[i].datetime
if delta_time.total_seconds() > time_thresh:
mean_point = np.mean([[p.lat, p.lon]
for p in traj[i:j + 1]],
axis=0)
arrival_time = traj[i].datetime
leave_time = traj[j].datetime
staypoints.append(
[mean_point, arrival_time, leave_time, i, j])
i = j
token = 1
break
j = j + 1
if not token == 1:
i = i + 1
return staypoints
def make_df(trajs):
data = list()
for traj in trajs:
if haversine_distance(traj[0], traj[-1]) < 100:
continue
fp = traj[0]
sps = staypoints_geolife(traj)
lp = traj[-1]
if len(sps) > 1:
data.append([
fp.lat, fp.lon, fp.datetime, sps[0][0][0], sps[0][0][1],
sps[0][1]
])
for i in range(1, len(sps) - 1):
data.append([
sps[i][0][0], sps[i][0][1], sps[i][1], sps[i + 1][0][0],
sps[i + 1][0][1], sps[i + 1][2]
])
data.append([
sps[-1][0][0], sps[-1][0][1], sps[-1][2], lp.lat, lp.lon,
lp.datetime
])
else:
data.append(
[fp.lat, fp.lon, fp.datetime, lp.lat, lp.lon, lp.datetime])
df = pd.DataFrame(data,
columns=[
'start_lat', 'start_lon', 'start_date', 'end_lat',
'end_lon', 'end_date'
])
df = df.set_index(pd.DatetimeIndex(df['start_date'])).sort_index()
return df
def resolve_endcluster(train, p):
end_clusters = {
Point(lat, lon, None): end
for lat, lon, end in zip(train['end_lat'], train['end_lon'],
train['end_cluster'])
}
end_cluster = None
min_dist = 400
for ec_point in end_clusters:
dist = haversine_distance(p.lat, p.lon, ec_point.lat, ec_point.lon)
if dist < min_dist:
min_dist = dist
end_cluster = end_clusters[ec_point]
return end_cluster