Parking lot map

As part of the autonomous valet parking usecase, we’ll need a map of a parking lot. It’d be great to have more than one parking lot to map, each with different characteristics (multi level, open air, underground, etc.) Parkopedia (@Brian_Holt) , Carmera and LG (@zelenkovsky) will work together to map and generate a suitable parking lot that can also be simulated in LG’s simulator.

Please use this thread to coordinate such efforts.

Hi Esteve,

We’ve started modeling the AutonomousStuff car park (off Hartog Drive) using the JOSM tooling just to get a feel for how it works and will contribute that to OSM as a matter of course. We’re also considering how the OSM XML format needs to change to support 3D (indoor facilities) and will propose some other car parks to map (we welcome suggestions).

@zelenkovsky Can we get access to the point cloud data that was collected for that facility?

Regards
Brian

Hello All,

Andrey K. has scanned the area around AutonomousStuff office already and we are processing it right now.

Mesh is 90% done, work on textures and UV maps is starting. We can share point cloud with image sequence, but it is ~50Gb. @Brian_Holt Would anybody be able to come to our office with HDD to save it? or your prefer to provide a place for uploading the data?

BR
Dmitry Z.

Hi Dmitry,

Is it a colourised point cloud? Could you send us a Dropbox or Google Drive link so that we can download the data?

Regards
Brian

Hello Brian,

This is not colorized point cloud, original ROS bag contains intensity, I’m not sure if produced point cloud still has it. But it includes series of images from the camera. We can share raw data and final 3D environment, I’m not sure any intermediate step is important for you.

Uploading to google drive could be a tedious work, I’ll try and see how long it takes.

Stay tuned.

BR
Dmitry Z.

Hi Dmity,

Have you had a chance to upload the file(s) to GDrive or Dropbox? We’d really like to be able to see them and include them in the modeling effort.

Regards
Brian

@zelenkovsky I can come by, pick up the data on the disk drive and take care of the upload. LMK. D.

@Dejan_Pangercic @Brian_Holt
What format do you prefer ? The original ROSBAG is in Apollo format, I guess you don’t want to decode Protobufs or yuyv images.
I have a combined and filtered (0.2m voxel filter) Point Cloud in PCD binary, individual (unfiltered) Lidar scans as PCDs + Lidar poses (UTM coordinates and quaternion), and undistorded images as jpgs + camera poses.

Hi Andrey,

My ideal would be an XYZRGB point cloud in something like a PCD for PLY format. While we could colourise the point cloud from images and extrinsics I’d prefer not to if we can avoid it. A point cloud with just XYZ is helpful for structure but it’s obviously difficult to identify lane markings etc.

Regards
Brian

Hi Brian,

Unfortunately we don’t colorize point cloud (yet) and currently it’s in XYZI, where Intensity is un-conventional int8

  • Andrey

Hi Andrey,

In that case we’ll take just the XYZI point cloud in PCD format. @Dejan_Pangercic could you work with @Andrey_Kozhanov to get this into the cloud so we can download it?

Thanks
Brian

@Brian_Holt no need for that, these can fit into my personal GD.
I’m uploading 2 files: full (but filtered) point cloud, and dense point cloud with road surface (~30 cm slice from ground level, +/- 15 meters from the Lidar in XY-plane)

https://drive.google.com/drive/folders/1m8cFeSzyAsc5Ue68AEhv-b7AZtWj53RN?usp=sharing

Let me know if you can access these or if you have any issues with the files themselves.
PCDs are rotated to the world frame (north-up), origin of the point cloud (0,0,0 point) is 596640.783512, 4137677.332274, -17.289531 in UTM or 37.380811523812845, -121.90840595108715 lat lon

1 Like