Trajectory Processing

traj_clean_drift(data[, col, method, ...])

Delete the drift in the trajectory data.

traj_clean_redundant(data[, col])

Delete the data with the same information as the data before and after to reduce the amount of data.

traj_slice(traj_data, slice_data[, ...])

Slice the trajectory data according to the slice data.

traj_smooth(data[, col, proj, ...])

Smooth Trajectory Using Kalman Filter.

traj_segment(data[, groupby_col, retain_col])

Segment the trajectory in order and return the starting and ending information of each segment.

traj_densify(data[, col, timegap])

Trajectory densification, ensure that there is a trajectory point each timegap seconds

traj_sparsify(data[, col, timegap, method])

Trajectory sparsify.

traj_stay_move(data, params[, col, activitytime])

Input trajectory data and gridding parameters, identify stay and move

traj_to_linestring(traj_points[, col, timecol])

Input trajectory, generate GeoDataFrame

traj_mapmatch(traj, G[, col])

Nearest map matching: Find the nearest point on the road network for each trajectory point.

traj_length(move_points[, col, method])

Calculate Trajectory Length.
_images/trajs.png
transbigdata.traj_clean_drift(data, col=['VehicleNum', 'Time', 'Lng', 'Lat'], method='twoside', speedlimit=80, dislimit=1000, anglelimit=30)

Delete the drift in the trajectory data. The drift is defined as the data with a speed greater than the speed limit or the distance between the current point and the next point is greater than the distance limit or the angle between the current point, the previous point, and the next point is smaller than the angle limit. The speed limit is 80km/h by default, and the distance limit is 1000m by default. The method of cleaning drift data is divided into two methods: ‘oneside’ and ‘twoside’. The ‘oneside’ method is to consider the speed of the current point and the next point, and the ‘twoside’ method is to consider the speed of the current point, the previous point, and the next point.

Parameters:

  • data (DataFrame) – Data

  • col (List) – Column names, in the order of [‘VehicleNum’, ‘Time’, ‘Lng’, ‘Lat’]

  • method (string) – Method of cleaning drift data, including ‘oneside’ and ‘twoside’

  • speedlimit (number) – Speed limitation

  • dislimit (number) – Distance limit

  • anglelimit (number) – Angle limit

Returns:

data1 – Cleaned data

Return type:

DataFrame

transbigdata.traj_clean_redundant(data, col=['VehicleNum', 'Time', 'Lng', 'Lat'])

Delete the data with the same information as the data before and after to reduce the amount of data. For example, if several consecutive data of an individual have the same information except for the time, only the first and last two data can be kept

Parameters:

  • data (DataFrame) – Data

  • col (List) – The column name, in the order of [‘Vehicleid, Time’]. It will sort by time, and then determine the information of other columns besides the time

Returns:

data1 – Cleaned data

Return type:

DataFrame

transbigdata.traj_slice(traj_data, slice_data, traj_col=['vid', 'time'], slice_col=['vid', 'stime', 'etime', 'tripid'])

Slice the trajectory data according to the slice data. This method extracts data from a given set of trajectory data(traj_data) based on a specified time period(slice_data).

Parameters:

  • traj_data (DataFrame) – Trajectory data, containing the trajectory of each vehicle

  • slice_data (DataFrame) – Slice data, containing the start time, end time and vehicleid of each slice

  • traj_col (List) – The column name of trajectory data, in the sequence of [VehicleNum, Time]

  • slice_col (List) – The column name of slice data, in the sequence of [VehicleNum_slice, Stime, Etime, SliceId]

Returns:

data_sliced – The sliced trajectory data

Return type:

DataFrame

Example

>>> tbd.traj_slice(GPSData, move, traj_col=['vid', 'time'], slice_col = ['vid','stime','etime','tripid'])
transbigdata.traj_smooth(data, col=['id', 'time', 'lon', 'lat'], proj=False, process_noise_std=0.5, measurement_noise_std=1)

Smooth Trajectory Using Kalman Filter.

Parameters:

  • data (DataFrame) – Trajectory data

  • col (list) – Column names of the trajectory data

  • proj (bool) – Whether to perform equidistant projection

  • process_noise_std (float) – Standard deviation of the process noise

  • measurement_noise_std (float) – Standard deviation of the measurement noise

Returns:

data – Smoothed trajectory data

Return type:

DataFrame

transbigdata.traj_segment(data, groupby_col=['id', 'moveid'], retain_col=['time', 'lon', 'lat'])

Segment the trajectory in order and return the starting and ending information of each segment. This function can segment GPS trajectory data, calculate the start and end information of each segment, and store the results in a DataFrame object. The input of this function includes a pandas DataFrame object containing GPS trajectory data, field names for grouping, and field names to be retained. The output is a pandas DataFrame object containing the starting and ending information of each segment, where each row represents a trajectory segment.

Parameters:

  • data (DataFrame) – The trajectory data needs to be sorted beforehand.

  • groupby_col (List) – A list of strings specifying the groupby fields to be used for segmentation.

  • retain_col (List) – A list of strings specifying the fields to be retained.

Returns:

data – Containing the starting and ending information of each segment, where each row represents a trajectory segment.

Return type:

DataFrame

transbigdata.traj_densify(data, col=['Vehicleid', 'Time', 'Lng', 'Lat'], timegap=15)

Trajectory densification, ensure that there is a trajectory point each timegap seconds

Parameters:

  • data (DataFrame) – Data

  • col (List) – The column name, in the sequence of [Vehicleid, Time, lng, lat]

  • timegap (number) – The sampling interval (second)

Returns:

data1 – The processed data

Return type:

DataFrame

transbigdata.traj_sparsify(data, col=['Vehicleid', 'Time', 'Lng', 'Lat'], timegap=15, method='subsample')

Trajectory sparsify. When the sampling frequency of trajectory data is too high, the amount of data is too large, which is not convenient for the analysis of some studies that require less data frequency. This function can expand the sampling interval and reduce the amount of data.

Parameters:

  • data (DataFrame) – Data

  • col (List) – The column name, in the sequence of [Vehicleid, Time, lng, lat]

  • timegap (number) – Time gap between trajectory point

  • method (str) – ‘interpolate’ or ‘subsample’

Returns:

data1 – Sparsified trajectory data

Return type:

DataFrame

transbigdata.traj_stay_move(data, params, col=['ID', 'dataTime', 'longitude', 'latitude'], activitytime=1800)

Input trajectory data and gridding parameters, identify stay and move

Parameters:

  • data (DataFrame) – trajectory data

  • params (List) – gridding parameters

  • col (List) – The column name, in the order of [‘ID’,’dataTime’,’longitude’, ‘latitude’]

  • activitytime (Number) – How much time to regard as activity

Returns:

  • stay (DataFrame) – stay information

  • move (DataFrame) – move information

transbigdata.traj_to_linestring(traj_points, col=['Lng', 'Lat', 'ID'], timecol=None)

Input trajectory, generate GeoDataFrame

Parameters:

  • traj_points (DataFrame) – trajectory data

  • col (List) – The column name, in the sequence of [lng, lat,trajectoryid]

  • timecol (str(Optional)) – Optional, the column name of the time column. If given, the geojson with [longitude, latitude, altitude, time] in returns can be put into the Kepler to visualize the trajectory

Returns:

traj – Generated trajectory

Return type:

GeoDataFrame

transbigdata.traj_mapmatch(traj, G, col=['lon', 'lat'])

Nearest map matching: Find the nearest point on the road network for each trajectory point. When conducting nearest neighbor matching, we need to find the closest road segment on the road network for each trajectory point, and match the trajectory point to that segment. In practice, we can first extract the nodes of the road segments to form a set of points (i.e., extracting each coordinate point from each LineString in the geometry column), then calculate the nearest distance between the trajectory point and this set of points, and finally match the trajectory point to the road segment where the nearest distance’s node is located. This process effectively transforms the problem of matching points to lines into a problem of matching points to points.

Parameters:

  • traj (DataFrame) – The trajectory point data set to be matched.

  • G (networkx multidigraph) – The road network used for matching, created by osmnx.

  • col (list) – The name of the longitude and latitude columns in the trajectory point data set.

Returns:

traj_matched – The trajectory point data set after matching.

Return type:

DataFrame

transbigdata.traj_length(move_points, col=['lon', 'lat', 'moveid'], method='Haversine')

Calculate Trajectory Length. Input the trajectory point data and calculate the length of each trajectory in meters.

Parameters:

  • move_points (DataFrame) – Trajectory point data, which includes trajectory id, longitude, latitude, etc. Different trajectories are distinguished by trajectory id, and trajectory points are arranged in time order.

  • col (list) – Column names of the trajectory point data, in the order of [longitude, latitude, trajectory id]

  • method (str) – The method of calculating the trajectory length, optional ‘Haversine’ or ‘Project’, default is ‘Haversine’ using the haversine formula to calculate spherical distance, ‘Project’ transforms the data into a projected coordinate system to calculate plane distance.

Returns:

move_trajs – Trajectory length data, including two columns of trajectory id and trajectory length, the unit is meters.

Return type:

DataFrame