Contact: | Francois Pomerleau |
---|---|
f.pomerleau [at] gmail [dot] com | |
Date: | Aug 25, 2011 |
Place: | UniversitätsSpital Park, Zurich, Switzerland |
Situation: | Outdoors |
Environments: | Unstructured (mainly vegetation) |
Dynamics: | People walking and seasonal changes |
Scanner path: | Slightly 3D before following a road |
Re-localization: | 0 |
Number of scan: | 37 |
Avg. points/scan: | 182 000 |
Path bounding box: | 10 x 15 x 0.5 m |
Scene bounding box: | 30 x 53 x 20 m |
This environment is mainly constituted of vegetation (tree, bushes, etc). The only structured element is a small paved road that crosses the wood. The scanner path starts in the wood and continues for approximately 12 scans before joining the small road (see Contextual Photographs).
The main motivation for this dataset is to evaluate robustness of registration algorithm to unstructured environment. Some people were walking on the road while recording. The complete representation can also be compared with Wood in autumn to highlight the global (seasonal) modification of the environment.
F. Pomerleau, M. Liu, F. Colas, and R. Siegwart, Challenging data sets for point cloud registration algorithms, International Journal of Robotic Research, vol. 31, no. 14, pp. 1705–1711, Dec. 2012. bibtex, Publisher Link
Manually drew map. The photo poses number correspond to the ones used in Contextual Photographs.
You can use right and left arrows keys to navigate once you click on a picture. Number in the photo titles correspond the those used in the Environment Topology section (click on the photo to see the numbers).
Top view of the wood with the upper part of the vegetation removed to ease visualization:
Scan sequence with the upper part of the vegetation removed to ease visualization:
Side view of the dataset with the small road (in red) passing through the vegetation:
Example of a seasonal change - a big tree with its leaves (point cloud manually extracted):
Overlap between each scan. The graph can be read as the percentage of points in Scan A that are also in Scan B. The overlap matrix can be downloaded in the section Download Point Clouds in Global Frame as a csv file.
All file contents and headers are explained on the page Tilting Laser - File Formats.
Point clouds of this section have their origin at the scanner center. 2D scans have been transformed using the axis encoder to produce consistent 3D point clouds. The supporting data (Gravity, Magnetic North and GPS) have been recorded during the time the scanner was rotating. If you do not wish to compute the statistics for the supporting data, you can go to the section Point Clouds in Global Coordinates were single measurements have been selected per pose. The “ground truth” poses can also be downloaded in the section Point Clouds in Global Coordinates.
All csv files have a header explaining each column and consistent timestamps.
Download all local files: local_frame.zip (155 MB) local_frame.tar.gz (155 MB)
- or -
Select the specific csv file you want here.
For Matlab users:
Point clouds of this section have been moved to a global reference frame where the pose of the first 3D scan is the origin. The supporting data (Gravity, Magnetic North and GPS) have been post-processed to have only one reading per 3D scan.
All csv files have a header explaining each column and consistent timestamps.
Download all global files: global_frame.zip (125 MB) global_frame.tar.gz (125 MB)
- or -
Select the specific csv file you want here or in the list below:
For Matlab users:
File type chosen for visualization is VTK legacy in ASCII format. We suggest to use Paraview to view the files because it can be freely downloaded for Windows, Mac and Ubuntu user and it's supported by Kitware. All screenshots of this page were realized using this software. You can download it here: http://www.paraview.org/
Download all local files: vtk_global.zip (124 MB) vtk_global.tar.gz (124 MB)
- or -
Select the specific vtk file you want here or in the table below:
This type of data can be useful if you want to do some preprocessing on the 2D scans directly. We used ROS
(www.ros.org) as middleware so the raw recordings can be downloaded and playback using rosbag
. One rosbag has been recored per 3D scan. The ground truth poses are not available in the rosbags
.
For more information on how to use this format see: www.ros.org/wiki/rosbag
Here is the output of rosbag info
for the first scan. All scans have roughly the same number of messages
path: 0-Tiltlaser.bagversion: 2.0 duration: 17.8s start: Aug 25 2011 13:06:59.18 (1314270419.18) end: Aug 25 2011 13:07:17.03 (1314270437.03) size: 1.4 MB messages: 4123 compression: none [2/2 chunks] types: geometry_msgs/TwistStamped [98d34b0043a2093cf9d9345ab6eef12e] geometry_msgs/Vector3Stamped [7b324c7325e683bf02a9b14b01090ec7] sensor_msgs/Imu [6a62c6daae103f4ff57a132d6f95cec2] sensor_msgs/LaserScan [90c7ef2dc6895d81024acba2ac42f369] sensor_msgs/NavSatFix [2d3a8cd499b9b4a0249fb98fd05cfa48] tf/tfMessage [94810edda583a504dfda3829e70d7eec] xsens_mtig/GPSInfoStatus2 [b43aadd9fb237c1b978f00a668605345] xsens_mtig/Thermistor [1a7c01d7f495652f5c21d90276c66498] topics: /gravity_vector 359 msgs : geometry_msgs/Vector3Stamped /imu_data 359 msgs : sensor_msgs/Imu /info_gps 359 msgs : xsens_mtig/GPSInfoStatus2 /magnetic_vector 359 msgs : geometry_msgs/Vector3Stamped /position_gps 359 msgs : sensor_msgs/NavSatFix /scan 179 msgs : sensor_msgs/LaserScan /temperature 718 msgs : xsens_mtig/Thermistor /tf 713 msgs : tf/tfMessage /velocity 718 msgs : geometry_msgs/TwistStamped
Download all bags files: rosbags.zip (18.4 MB), rosbags.tar.gz (18.3 MB)
- or -
Select the specific bag you want here.