def plot_users_trail(users, signals, cells, save_pic=False, save_trails_js=False):
'''
save users' trail plot or save users' trail js file
:param users: users id array
:param signals: signals
:param cells: get_cells
:param save_pic: Boolean Whether to save the trail plot
:param save_trails_js: Boolean Whether to save the trail js file
:return:
'''
n = len(users)
for i in range(n):
if save_pic & save_trails_js:
break
user_id = users[i]
signal = signals[signals.user_id == user_id]
signal = signal.sort_values('dates')
signal = pd.merge(signal, cells, on='cell_id', how='left')
signal = signal.dropna(0)
print('user %d' % i)
if save_pic:
plt.plot(signal.lon, signal.lat)
plt.savefig('../../pic/%.2d%.2d/%.2dt%.2d/pic%d.jpg' % (mon, dd, hh_start, hh_end, i))
plt.close('all')
if save_trails_js:
output_data_js(signal, '../../web/data/%.2d%.2d/%.2dt%.2d/user%d.js' % (mon, dd, hh_start, hh_end, i))
def output_pre_road_data(trail,
direction,
n,
type='cell',
interpolate_interval=1,
smoothing=0,
penalty=0.9,
direction_window=15):
'''
Output predicted road data with CTrack. For test wuhu data.
:param trail: <Int> The i_th trail
:param direction: <-1/1> -1 means reverse side; 1 means front.
:param n: The n_th test of this trail. If None, return all the tests of the trail.
:param type: 'cell' or 'GPS'. See details in function get_obs_data.
:param interpolate_interval: Interval seconds of processed observables. Default = 1.
:param smoothing: Window length of smoothing. If smoothing == 0, do not smooth.
:param penalty:
:param direction_window:
:return: None.
'''
df = get_road_tests(trail, direction, n)
graph = get_graph('../../data/road_test/link_baidu_wuhu.txt')
obs = get_obs_data(df, type=type)
track = CTrack(obs,
graph,
smoothing=smoothing,
interpolate_interval=interpolate_interval,
penalty=penalty,
direction_window=direction_window)
states = track.viterbi()
direction = u'正向' if direction == 1 else u'反向'
output_data_js(
states,
'../../res/MapMatching_test/pre_%s/wuhu_pre_线路%d_%s%d_smoothing=%d_window=%d.js'
% (type, trail, direction, n, smoothing, interpolate_interval))
def plot_users_trail(file_date,
save_pic=True,
save_trails_js=True,
save_data=True,
save_dir=work_dir + 'attach/'):
'''
save users' trail plot or save users' trail js file
:param users: users id array
:param signals: signals
:param cells: get_cells
:param save_pic: Boolean Whether to save the trail plot
:param save_trails_js: Boolean Whether to save the trail js file
:return:
'''
cells = pd.read_csv('../../res/[0926]cells_process.csv')
cells_loc = cells[['cell_id', 'lat', 'lon']]
dirs = [save_dir + 'user_pic/', save_dir + 'user_trail/']
signals = get_signals_from_csv(save_dir + 'user_data' + file_date + '.csv')
labels = list(pd.read_csv(save_dir + 'user_pic/random_user.csv').user_id)
users = [x.split('@')[0] for x in labels]
n = len(users)
for i in range(n):
# if save_pic & save_trails_js:
# break
user_id = users[i]
signal = signals[signals.user_id == user_id]
signal = signal.sort_values('dates')
signal = pd.merge(signal, cells, on='cell_id', how='left')
signal = signal.dropna(0)
print('user %d' % i)
if save_pic:
plt.plot(signal.lon, signal.lat)
plt.savefig('%suser_trail_%s/pic/user%d.jpg' %
(save_dir, file_date, i))
plt.close('all')
if save_data:
signal.to_csv('%suser_trail_%s/csv/user%d.csv' %
(save_dir, file_date, i),
index=False)
if save_trails_js:
output_data_js(
signal,
'%suser_trail_%s/data/user%d.js' % (save_dir, file_date, i))
def output_obs_data(trail,
direction,
n,
type='cell',
interpolate_interval=5,
smoothing=20,
interpolate=True):
'''
Output observables data after pre-processing. For test wuhu data
:param trail: <Int> The i_th trail
:param direction: <-1/1> -1 means reverse side; 1 means front.
:param n: The n_th test of this trail. If None, return all the tests of the trail.
:param type: 'cell' or 'GPS'. See details in function get_obs_data.
:param interpolate_interval: Interval seconds of processed observables. Default = 1.
:param smoothing: Window length of smoothing. If smoothing == 0, do not smooth.
:param interpolate:
:return: None.
'''
df = get_road_tests(trail, direction, n)
graph = get_graph('../../data/road_test/link_baidu_wuhu.txt')
obs = get_obs_data(df, type=type)
track = CTrack(obs,
graph,
smoothing=smoothing,
interpolate_interval=interpolate_interval,
interpolate=interpolate)
out = track.obs
out['dates'] = out.time
direction = u'正向' if direction == 1 else u'反向'
if interpolate:
output_data_js(
out,
'../../res/MapMatching_test/processed_%s/wuhu_线路%d_%s%d_smoothing=%d_window=%d.js'
% (type, trail, direction, n, smoothing, interpolate_interval))
else:
output_data_js(
out, '../../res/MapMatching_test/raw_%s/wuhu_线路%d_%s%d.js' %
(type, trail, direction, n))
def plot_users_trail(self,
users=None,
within_users=True,
save_pic=False,
save_trails_js=False):
'''
save users' trail plot or save users' trail js file
:param users: users id array. If users is None, it will plot all the users. Default = None
:param within_users: read signals within users or not. Default = True
:param save_pic: Boolean Whether to save the trail plot
:param save_trails_js: Boolean Whether to save the trail js file
:return:
'''
cells = pd.read_csv('../../res/[0926]cells_process.csv')
signals = self.get_clean_signals(within_users)
if users is None:
users = signals.user_id.unique()
n = len(users)
for i in range(n):
if not (save_pic | save_trails_js):
break
user_id = users[i]
signal = signals[signals.user_id == user_id]
signal = signal.sort_values('dates')
signal = pd.merge(signal, cells, on='cell_id', how='left')
signal = signal.dropna(0)
print('user %d' % i)
if save_pic:
self.check_dir(self.plot_dir)
plt.plot(signal.lon, signal.lat)
plt.title('user = %s' % user_id)
plt.savefig('%s/pic%d.jpg' % (self.plot_dir, i))
plt.close('all')
if save_trails_js:
self.check_dir(self.js_dir)
output_data_js(signal, user_id,
'%s/user%d.js' % (self.js_dir, i))
def single_map_matching(obs,
graph,
save_name,
mm_smoothing=10,
mm_interpolate_interval=10,
save_dir='../../res/work/0417_multi_mm/'):
output_data_js(obs, save_dir + save_name + '_raw.js')
track = CTrack(obs,
graph,
smoothing=mm_smoothing,
interpolate_interval=mm_interpolate_interval)
output_data_js(track.obs, save_dir + save_name + '_processed.js')
states = track.viterbi()
output_data_js(states, save_dir + save_name + '.js')
return states
def multi_map_matching(obses,
graph,
save_name,
mdtw_interval=5,
mm_smoothing=10,
mm_interpolate_interval=10,
save_dir='../../res/work/0417_multi_mm/'):
mdtw = MultiDTWIndx(obses, interval=mdtw_interval)
new_obs = mdtw.run_one_by_one()
print(new_obs)
output_data_js(new_obs, save_dir + save_name + '_raw.js')
track = CTrack(new_obs,
graph,
smoothing=mm_smoothing,
interpolate_interval=mm_interpolate_interval)
output_data_js(track.obs, save_dir + save_name + '_processed.js')
states = track.viterbi()
output_data_js(states, save_dir + save_name + '.js')
return states
graph = get_graph('../../data/hefei_road/link_baidu.txt')
user_id = users[10]
signal, ind_sets, signal_sets = get_trajectory(signals, cells_loc, user_id)
print(cal_rough_speed_per_cell2(signal))
for i in range(len(signal_sets)):
if i != 1: continue
obs = pd.DataFrame({
'time': signal_sets[i]['dates'],
'lon': signal_sets[i]['lon'],
'lat': signal_sets[i]['lat']
})
# print(obs)
obs = deal_same_obs(obs)[0:8]
check_directory('../../res/MapMatching/%s' % user_id)
output_data_js(
obs, '../../res/MapMatching/%s/raw_trajectory%d.js' % (user_id, i))
obs = deal_outlier(obs)
speed = cal_rough_speed_per_cell2(obs)
print(speed)
print([
abs(speed['direction'][i] - speed['direction'][i - 1])
for i in range(1, len(speed))
])
track = CTrack(obs, graph, smoothing=20, interpolate_interval=5)
out = track.obs
states = track.viterbi()
track.plot_each_time_best_trajectory(track.delta)
# print(cal_rough_speed_per_cell2(states))
output_data_js(
out, '../../res/MapMatching/%s/pre_trajectory%d.js' % (user_id, i))
output_data_js(