Note

This page was generated from user_guide/data_visualization/trajectory.pct.py.

Interactive online version:

Visualization of trajectories

If we work with trajectory data, we often want to visualize them from a so called birds eye view. The following example demonstrates how to achieve this with TASI using the DLR Urban Traffic Dataset.

Load trajectories

At first, let’s load trajectories from the DLR dataset.

from tasi.dlr import DLRTrajectoryDataset, DLRUTDatasetManager, DLRUTVersion

dataset = DLRUTDatasetManager(DLRUTVersion.latest)
dataset.load()
ut = DLRTrajectoryDataset.from_csv(dataset.trajectory()[0])
ut

Downloading DLR-Urban-Traffic-dataset_v1-3-0: 100%|██████████| 420M/420M [05:50<00:00, 1.20MB/s]
Extracting: 100%|██████████| 592/592 [00:19<00:00, 30.10file/s]

		acceleration			position		classifications						dimension			interpolated	velocity			yaw
		easting	magnitude	northing	easting	northing	bicycle	car	motorbike	pedestrian	truck	van	height	length	width		easting	magnitude	northing
timestamp	id
2023-09-24 00:00:00.016482+00:00	1695513598769889	0.047	0.093	0.080	604825.208	5.793e+06	0.0	0.645	0.205	0.0	0.054	0.096	1.562	4.294	1.811	False	-12.789	13.496	-4.311	-161.405
2023-09-24 00:00:00.066482+00:00	1695513598769889	0.052	0.097	0.081	604824.540	5.793e+06	0.0	0.645	0.205	0.0	0.054	0.096	1.562	4.294	1.811	False	-12.788	13.494	-4.308	-161.411
2023-09-24 00:00:00.116482+00:00	1695513598769889	0.058	0.101	0.083	604823.871	5.793e+06	0.0	0.645	0.205	0.0	0.054	0.096	1.562	4.294	1.811	False	-12.787	13.493	-4.305	-161.418
2023-09-24 00:00:00.166482+00:00	1695513598769889	0.063	0.105	0.084	604823.202	5.793e+06	0.0	0.645	0.205	0.0	0.054	0.096	1.562	4.294	1.811	False	-12.786	13.490	-4.302	-161.426
2023-09-24 00:00:00.216482+00:00	1695513598769889	0.069	0.109	0.085	604822.533	5.793e+06	0.0	0.645	0.205	0.0	0.054	0.096	1.562	4.294	1.811	False	-12.785	13.488	-4.298	-161.435
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2023-09-24 00:14:59.766482+00:00	1695514432518548	-0.006	0.013	-0.012	604743.005	5.793e+06	0.0	0.135	0.865	0.0	0.000	0.000	1.656	1.058	0.675	False	-0.021	0.022	-0.006	8.619
2023-09-24 00:14:59.816482+00:00	1695514432518548	-0.005	0.012	-0.011	604743.003	5.793e+06	0.0	0.135	0.865	0.0	0.000	0.000	1.656	1.058	0.675	False	-0.022	0.023	-0.008	8.619
2023-09-24 00:14:59.866482+00:00	1695514432518548	-0.004	0.011	-0.010	604743.001	5.793e+06	0.0	0.135	0.865	0.0	0.000	0.000	1.656	1.058	0.675	False	-0.022	0.025	-0.011	8.619
2023-09-24 00:14:59.916482+00:00	1695514432518548	-0.004	0.009	-0.009	604742.999	5.793e+06	0.0	0.135	0.865	0.0	0.000	0.000	1.656	1.058	0.675	False	-0.023	0.027	-0.013	8.619
2023-09-24 00:14:59.966482+00:00	1695514432518548	-0.003	0.008	-0.007	604742.996	5.793e+06	0.0	0.135	0.865	0.0	0.000	0.000	1.656	1.058	0.675	False	-0.024	0.028	-0.015	8.619

33614 rows × 19 columns

Plot trajectories

We now utilize the TrajectoryPlotter to visualize the trajectories of the traffic participants that we’ve loaded.

import matplotlib.pyplot as plt

from tasi.plotting import TrajectoryPlotter

f, ax = plt.subplots()

plotter = TrajectoryPlotter()
plotter.plot(ut, ax=ax)

Change trajectory color

Note that the trajectories are colorized with the default color which we can change so any arbitrary color, such as lightgray.

f, ax = plt.subplots()
plotter.plot(ut, ax=ax, color="lightgray")

Change color of a specific trajectory

You can also define the color of a specific trajectory.

f, ax = plt.subplots()
plotter.plot(
    ut, ax=ax, color="black", trajectory_kwargs={ut.ids[-20]: {"color": "red"}}
)

Change trajectory opacity

or even change the opacity of each trajectory to quickly get an overview of the traffic density. Let’s change the opacity to 20% via the alpha argument.

f, ax = plt.subplots()
plotter.plot(ut, ax=ax, alpha=0.2)

The plot already indicates the various traffic volumes on the different routes at the intersection.

Plot DLR UT trajectories on orthophoto

We can also combine the TrajectoryPlotter with the BoundingboxPlotter to visualize trajectories on an orthophoto.

from tasi.plotting.wms import BoundingboxPlotter, LowerSaxonyOrthophotoTile

f, ax = plt.subplots()

# plot the orthophoto first
bbox_plotter = BoundingboxPlotter(ut.roi, LowerSaxonyOrthophotoTile())
bbox_plotter.plot(ax)

# and the trajectories on top
tj_plotter = TrajectoryPlotter()
tj_plotter.plot(ut, ax=ax)

# hide the axis
_ = ax.axis("off")

Note that it may become hard to see the trajectories on the background. To increase the contrast, we can adapt the opacity of the ortophoto via the alpha argument. Let’s set is to 0.5 (50%) to highlight the trajectories.

f, ax = plt.subplots()

bbox_plotter.plot(ax, alpha=0.5)
tj_plotter.plot(ut, ax=ax, alpha=0.25)

_ = ax.axis("off")

Plot DLR HT trajectories on orthophoto

We can also use the TrajectoryPlotter with the BoundingboxPlotter to visualize trajectories of the DLR HT dataset on an orthophoto. Let’s load and plot the trajectories from the DLR HT dataset. This time, let’s use the DLRTrajectoryPlotter to color the trajectories in default DLR-colors.

from tasi.dlr.dataset import DLRHTDatasetManager, DLRHTVersion
from tasi.dlr.plotting import DLRTrajectoryPlotter

# load dataset
dataset = DLRHTDatasetManager(DLRHTVersion.v1_1_0)
dataset.load()
ht = DLRTrajectoryDataset.from_csv(dataset.trajectory()[0])

f, ax = plt.subplots()

# plot the orthophoto first
bbox_plotter = BoundingboxPlotter(ht.roi, LowerSaxonyOrthophotoTile())
bbox_plotter.plot(ax)

# and the trajectories on top
tj_plotter = DLRTrajectoryPlotter()
tj_plotter.plot(ht, ax=ax)

# hide the axis
_ = ax.axis("off")

Visualizing geospatial trajectories

The TASI model for representing trajectory data using geospatial objects opens the world to utilize tools that support geopandas. For instance, if you want an interactive view on trajectory data, we can utilize folium via geopandas. For this purpose, we need two conversion steps.

At first, we need to convert the dataset to a native TASI representation.

ds = ut.to_tasi()
ds.head()

		position		velocity			acceleration			heading	yaw_rate	...	boundingbox
		easting	northing	easting	magnitude	northing	easting	magnitude	northing			...	front		front_right		right		rear_right		rear
												...	easting	northing	easting	northing	easting	northing	easting	northing	easting	northing
timestamp	id
2023-09-24 00:00:00.016482+00:00	1695513598769889	604825.208	5.793e+06	-12.789	13.496	-4.311	0.047	0.093	0.080	-2.817	NaN	...	604827.352	5.793e+06	604827.308	5.793e+06	604825.163	5.793e+06	604823.019	5.793e+06	604823.064	5.793e+06
2023-09-24 00:00:00.066482+00:00	1695513598769889	604824.540	5.793e+06	-12.788	13.494	-4.308	0.052	0.097	0.081	-2.817	-0.002	...	604826.684	5.793e+06	604826.640	5.793e+06	604824.495	5.793e+06	604822.351	5.793e+06	604822.396	5.793e+06
2023-09-24 00:00:00.116482+00:00	1695513598769889	604823.871	5.793e+06	-12.787	13.493	-4.305	0.058	0.101	0.083	-2.817	-0.002	...	604826.015	5.793e+06	604825.971	5.793e+06	604823.826	5.793e+06	604821.682	5.793e+06	604821.727	5.793e+06
2023-09-24 00:00:00.166482+00:00	1695513598769889	604823.202	5.793e+06	-12.786	13.490	-4.302	0.063	0.105	0.084	-2.817	-0.003	...	604825.346	5.793e+06	604825.302	5.793e+06	604823.157	5.793e+06	604821.013	5.793e+06	604821.058	5.793e+06
2023-09-24 00:00:00.216482+00:00	1695513598769889	604822.533	5.793e+06	-12.785	13.488	-4.298	0.069	0.109	0.085	-2.818	-0.003	...	604824.677	5.793e+06	604824.633	5.793e+06	604822.488	5.793e+06	604820.344	5.793e+06	604820.389	5.793e+06

5 rows × 35 columns

Note that the boundingbox attribute was added in this step. Since ds is a tasi.TrajectoryDataset it also uses pandas. Hence, we will now convert to use geopandas while the position attribute should be encoded as a GeoObject.

gds = ds.as_geo("position")

We need to set the position attribute as the active geometry.

gds.set_geometry("position", inplace=True)

Now, we are ready to visualize the trajectories in a dynamic window, while we use the “CartoDB positron” background layer for reference.

gds.explore(crs="EPSG:32632", tiles="CartoDB positron")

Make this Notebook Trusted to load map: File -> Trust Notebook

Note that for each traffic participant (or trajectory), additional information is available when hovering over its representation in the map. You can customize the representaton of the trajectories and add or remove attributes to you liking.