If DMS has a need for the original point cloud data please let me know. It might be helpful for planning, modifications to the building, laying out committee areas, etc. The original data is around 390 million points and includes the ceilings that were removed for the renders. The renders are 2160P, I recommend opening them in a new tab. 
35 Likes
Who’s that in the unisex bathroom?
Dunno, I didn’t go into bathrooms. Some of the data points are artifacts like from mirrors or really shiny surfaces.
I was just being stoopid/annoying.
Damn cool technology and pics, though.
2 Likes
Really cool! Do you have more info about the process/equipment used to generate the renders?
1 Like
The LIDAR scan data came from a GeoSLAM ZEB Horizon scanner which is based on the Velodyne Puck.
https://geoslam.com/solutions/zeb-horizon/
The software I used to render the data is called Houdini which is an incredibly powerful and flexible computer graphics program.
10 Likes
This is seriously cool!
3 Likes
Absolutely awesome!
This is soooo cool. Could we like, take this model and throw it on a touchscreen monitor and make an interactive map of the space?
3 Likes
That’s a possibility. Unreal Engine has a plug-in to work with LIDAR point clouds.
4 Likes
How does it create the shadows like the dust collector pipes in woodshop?
Is that added when it is rendered.
The shadows (pipes, walls) appear to be a post-processing feature.
2 Likes
How tough would it be to get a 3D-printable STL?
2 Likes
The renderer is adding the lighting in this case.
A mesh would be doable but making it 3D printable would be pretty difficult. Without lots of processing it would have what amounts to thousands of delicate overhanging pieces that wouldn’t be practical to print.
I have heard that there are tools that can attempt to translate a cloud of point data into more regular finite geometric objects. Assuming the goal here is something that merely looks right as opposed to holding a machining tolerance seems like one of those tools would satisfy.
2 Likes
For sure! There’s lots of interesting stuff emerging constantly and identification of objects based on point clouds is a thing. What really complicates the task is making the model 3D printable. There are so many gotchas with 3D printing.
The low hanging fruit solution would be to cast rays from top to bottom and only include things that are in the first collision of the ray. So a complex object like a table becomes a box which makes it easy to 3D print because all of the overhangs are gone. However things like ducts hanging from the ceiling would become walls because the first collision is basically at the ceiling.
Creating a 3D printable model is doable but it would take more hours of labor than I’m willing to do on a whim (I’ll send the original data if someone wants to try). At some point in the future there might be a filter or a button to make this happen but we’re not there yet.
2 Likes
Overhang would be a challenge for 3D printing a floor plan, for sure.
1 Like
This is so cool!!! I had my '72 Opel GT in the second bay in automotive when you came through with the LIDAR. I would really love to have the point cloud data from that part of the scan… or all of it, really, to play around with getting it into something like Unity or Unreal. And to convert the scan of the GT into an editable model with which I can have much fun and create things like an air dam for the front end. It’s the ultimate 3D scanner!
And @Bizwacky, that’s a great idea for the interactive map! I’ve been planning to create an interactive map app for the space that’s in two parts (as soon as @jast gets my workspace set up on the DMS servers); an app that will geolocate where you are and provide a way to select where you’d like to go and show you the paths to get there, and a secondary app or another aspect of the map app that lets you search for and locate specific tools, accessories, supplies, etc. The LIDAR map would make it that much more interesting.
I now, very stupidly, want to make a VR app that could be used for virtual tours… friggin’ masochist, evidently… 
Thanks for doing this and posting it, @lukeiamyourfather! Very much appreciated!
2 Likes
I would greatly appreciate a copy of the original data to play with, if you would be so kind.
Thanks very much!
The accuracy of this setup is about 1 cm so it’s probably not good enough to create functional parts from. Accuracy of 1 cm at the scale of a city block is great. Accuracy of 1 cm on a 1 m part isn’t so good. Other 3D scanning methods like laser triangulation, structured light, or photogrammetry would be better for this because they have much better accuracy (sub mm in some cases). I think DMS has all three of those but it’s been a while so not sure at this point. Photo is from a class about laser triangulation scanner that DMS had/has.
I exported it with and without the ceilings and in 3 formats. None are native to LIDAR because the tool I’m using isn’t really a LIDAR tool. They can be converted to whatever formats though (via MeshLab, CloudCompare, Blender, etc.). The original LAZ files aren’t aligned with each other and they’re missing some info like normals and time stamps so these should be more useful than the original data.
BGEO - Binary format native to Houdini
PLY - ASCII based format that is widely supported
ABC - Binary format that is widely supported
They’ll be in my member folder on the file server when I figure out how to get them there which will probably be later today. It’s fewer points than I thought at 369 million points but the files are still pretty big (PLY file is 31 GB).
2 Likes