Convert to global system coordinate

[dandon223]

If I understand correctly both 3D annotations and lidar point clouds are in lidar system coordinate. Will I be able with this dataset to transform both of them to global system coordinate for example to do aggregated point cloud? For example you can do something like that with Nuscenes dataset and I wondered if here it is also possible.

[TamasMatuszka]

Hi @dandon223,

Sure, you can build an aggregated point cloud. LiDAR points are in the body coordinate system, and we used ECEF as the global coordinate system. You can find more details about the coordinate systems in Section 3.2 of https://openreview.net/pdf?id=LW3bRLlY-SA.

The ego-motion (sensor/gnssins/egomotion2.json) can be used to transform the points into an absolute coordinate system.

egomotion.json parameters:

• key: camera frame ID
• value: RT matrix from body coordinate system to ECEF

egmotion2.json parameters:

• key: camera frame id (like in egomotion.json), value is a dictionary where:
• RT_ECEF_body is the same transformation matrix as in egomotion.json
• enh_sep is the latitude, longitude, and height separation, i.e., GNSS error
• rph_sep is the roll, pitch, and heading separation, i.e., IMU error
• time: GNSS+INS device timestamp
• time_host: synchronized host timestamp (from camera metadata)

You can also use this file for calculating the transformation between frames.

Pull requests are welcome. If you build such a feature, we are happy to merge it :)

[TamasMatuszka]

Hi @dandon223,

We updated the manuscript with dataset documentation, you can find it here. This might also contain some useful information for you.