Point cloud formats, structured/unstructured

To chat about anything else.
VXGrid
V.I.P Member
V.I.P Member
Posts: 230
Joined: Fri Feb 24, 2017 10:47 am
Full Name: Martin Graner
Company Details: PointCab GmbH
Company Position Title: Research and Development
Country: Germany
Linkedin Profile: No
Has thanked: 51 times
Been thanked: 56 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by VXGrid »

Peyman Bashiri wrote: Wed Apr 28, 2021 9:20 pm
smacl wrote: Wed Apr 28, 2021 8:38 pm
Peyman Bashiri wrote: Wed Apr 28, 2021 7:34 pm I have a question too, is it possible to convert/export unstructured point cloud to structured?
Peyman
If you had the position and orientation details for each setup, along with a way to link them back to the related scan points, yes.
Thanks, I was asking about aerial LiDAR (drone or fixed wing).
Peyman
Shane beat me to it while I was writing my novel :)
I'll post it anyway

I think here we need to define what structured means.
Terrestrial scan data is, when it comes from the scanner, in the first step structured. Every captured point corresponds to the scannerhead, which is the Origin, since the scanner produces local coordinates. Thus we can use that point cloud data to visualize it as a 360 Panorama.
Now going to a mobile scanner: Normally the data is not taken stationary, meaning the captured points correspond to different capture positions.
If we go into images as an example: If you take a 360 Panorama, your camera is at the same position (terrestrial), in contrast when taking a video, while moving around, the position of the camera is different for every picture in the video stream.

Aerial LiDAR data is normally taken as a stripe, rather than an image (only one scan line, not a complete scan frame), so one shot of the scan line corresponds to one position of the drone.
There is relation between where the points are taken and the taken points, but you can't visualize them as a Panorama.

Which means: One can argue, that this point cloud data is structured, as long as there is knowledge about where the captured points (from which position and view direction) were taken. So you need knowledge about the trajectory (position + view direction) + the relation between cloud data and trajectory.

The other question which comes to mind: Is it possible to change mobile captured data into terrestrial 360° panorama data?
Answer is: Yes, depending on the used algorithm it will work better or worse, but I think I have not seen anybody doing that fully automatic (like choosing the virtual scan positions, so every point in the cloud is used in the positions).
User avatar
Daniel Wujanz
V.I.P Member
V.I.P Member
Posts: 124
Joined: Fri Aug 24, 2018 11:26 am
Full Name: Daniel Wujanz
Company Details: technet GmbH
Company Position Title: 3D Laser Scanning Specialist
Country: Germany
Linkedin Profile: Yes
Location: Berlin
Has thanked: 83 times
Been thanked: 55 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by Daniel Wujanz »

VXGrid wrote: Thu Apr 29, 2021 7:51 am
Peyman Bashiri wrote: Wed Apr 28, 2021 9:20 pm
smacl wrote: Wed Apr 28, 2021 8:38 pm

If you had the position and orientation details for each setup, along with a way to link them back to the related scan points, yes.
Thanks, I was asking about aerial LiDAR (drone or fixed wing).
Peyman
Shane beat me to it while I was writing my novel :)
I'll post it anyway

...

The other question which comes to mind: Is it possible to change mobile captured data into terrestrial 360° panorama data?
Answer is: Yes, depending on the used algorithm it will work better or worse, but I think I have not seen anybody doing that fully automatic (like choosing the virtual scan positions, so every point in the cloud is used in the positions).
Same here - Shane also spanked my arse whilst I was preparing the trilogy of a response : )

Anyway, another vital piece of information apart from the point cloud and the trajectory are time stamps for both data sources. The sum of data allows you to retrieve the information from where a point was taken and at which point in time.

Every kinematic system has to deal with temporal information. Unfortunately, the curse of the laser scanning business strikes again - the information is there - yet, there's (quite often) no way to retrieve it. The following manufacturers allow to access this data (this is an inclomplete list and will be extended):

- Kaarta
- DotProduct
- GeoSlam
- Leica Pegasus2 (but not BLK2GO @Paul: Please note that this is not intended to be Leica-bashing - it is just not understandable from a client's point of view)
- Z+F Profiler series

What Martin described is exactly the way we do it - we take the entire helix / kinematic point cloud / point sausage and cut it into "static" pieces. It is important to have no occlusions within a section. Interesting subject - I'll discuss it in greater detail in the PointCab webinar series in roughly a month.

All the best

Daniel
User avatar
smacl
Global Moderator
Global Moderator
Posts: 621
Joined: Tue Jan 25, 2011 5:12 pm
Full Name: Shane MacLaughlin
Company Details: Atlas Computers Ltd
Company Position Title: Managing Director
Country: Ireland
Linkedin Profile: Yes
Location: Ireland
Has thanked: 204 times
Been thanked: 198 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by smacl »

VXGrid wrote: Thu Apr 29, 2021 7:51 amShane beat me to it while I was writing my novel :)
I'll post it anyway

I think here we need to define what structured means.
Terrestrial scan data is, when it comes from the scanner, in the first step structured. Every captured point corresponds to the scannerhead, which is the Origin, since the scanner produces local coordinates. Thus we can use that point cloud data to visualize it as a 360 Panorama.
Now going to a mobile scanner: Normally the data is not taken stationary, meaning the captured points correspond to different capture positions.
If we go into images as an example: If you take a 360 Panorama, your camera is at the same position (terrestrial), in contrast when taking a video, while moving around, the position of the camera is different for every picture in the video stream.

Aerial LiDAR data is normally taken as a stripe, rather than an image (only one scan line, not a complete scan frame), so one shot of the scan line corresponds to one position of the drone.
There is relation between where the points are taken and the taken points, but you can't visualize them as a Panorama.

Which means: One can argue, that this point cloud data is structured, as long as there is knowledge about where the captured points (from which position and view direction) were taken. So you need knowledge about the trajectory (position + view direction) + the relation between cloud data and trajectory.

The other question which comes to mind: Is it possible to change mobile captured data into terrestrial 360° panorama data?
Answer is: Yes, depending on the used algorithm it will work better or worse, but I think I have not seen anybody doing that fully automatic (like choosing the virtual scan positions, so every point in the cloud is used in the positions).
Looking forward to the novel :)

Without being too pedantic, I think static and mobile scanning are better terms than terrestrial and aerial LIDAR. e.g. Pegasus back back, Geoslam, BLK2Go are terrestrial mobile scanners. Single scan line systems are usually referred to as profilers, e.g. Pegasus 2 and Pegasus ultimate, as opposed to multihead or interlaced systems such as Topcon and Pegasus Swift / Pegasus to go. You also get dual profilers like the Riegl mobile mapping solutions,
User avatar
smacl
Global Moderator
Global Moderator
Posts: 621
Joined: Tue Jan 25, 2011 5:12 pm
Full Name: Shane MacLaughlin
Company Details: Atlas Computers Ltd
Company Position Title: Managing Director
Country: Ireland
Linkedin Profile: Yes
Location: Ireland
Has thanked: 204 times
Been thanked: 198 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by smacl »

VXGrid wrote: Wed Apr 28, 2021 4:16 pmI'm preparing a webinar (it's upcoming Tuesday...) "Structured vs. unstructured point clouds"
And just signed up, looking forward to it.
VXGrid
V.I.P Member
V.I.P Member
Posts: 230
Joined: Fri Feb 24, 2017 10:47 am
Full Name: Martin Graner
Company Details: PointCab GmbH
Company Position Title: Research and Development
Country: Germany
Linkedin Profile: No
Has thanked: 51 times
Been thanked: 56 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by VXGrid »

smacl wrote: Thu Apr 29, 2021 12:42 pm
VXGrid wrote: Wed Apr 28, 2021 4:16 pmI'm preparing a webinar (it's upcoming Tuesday...) "Structured vs. unstructured point clouds"
And just signed up, looking forward to it.
:o
Panic mode enabled!
User avatar
smacl
Global Moderator
Global Moderator
Posts: 621
Joined: Tue Jan 25, 2011 5:12 pm
Full Name: Shane MacLaughlin
Company Details: Atlas Computers Ltd
Company Position Title: Managing Director
Country: Ireland
Linkedin Profile: Yes
Location: Ireland
Has thanked: 204 times
Been thanked: 198 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by smacl »

VXGrid wrote: Thu Apr 29, 2021 1:31 pmPanic mode enabled!
Good webinar guys, very well presented with no panic whatsoever :)
User avatar
landmeterbeuckx
V.I.P Member
V.I.P Member
Posts: 864
Joined: Tue May 01, 2012 5:19 pm
Full Name: Lieven Beuckx
Company Details: Studiebureau Beuckx
Company Position Title: Owner
Country: Belgium
Linkedin Profile: Yes
Has thanked: 68 times
Been thanked: 175 times

Re: Point cloud formats, structured/unstructured

Post by landmeterbeuckx »

Followed a portion of it and thought it was quite interesting. When the recording becomes available i'll see the whole thing through.
LSBbvba
Surveying services - 3D Laserscanning
Tel : +32477753126
www.lsbbvba.be
[email protected]
VXGrid
V.I.P Member
V.I.P Member
Posts: 230
Joined: Fri Feb 24, 2017 10:47 am
Full Name: Martin Graner
Company Details: PointCab GmbH
Company Position Title: Research and Development
Country: Germany
Linkedin Profile: No
Has thanked: 51 times
Been thanked: 56 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by VXGrid »

landmeterbeuckx wrote: Wed May 05, 2021 3:04 pm Followed a portion of it and thought it was quite interesting. When the recording becomes available i'll see the whole thing through.
A link to the recording should be in the follow up email, which we send normally around 24 hours after the webinar, so you should have the link by now, or in a couple of minutes.

If Marketing decides to upload the video to our youtube channel, I will share a link to that here.
User avatar
smacl
Global Moderator
Global Moderator
Posts: 621
Joined: Tue Jan 25, 2011 5:12 pm
Full Name: Shane MacLaughlin
Company Details: Atlas Computers Ltd
Company Position Title: Managing Director
Country: Ireland
Linkedin Profile: Yes
Location: Ireland
Has thanked: 204 times
Been thanked: 198 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by smacl »

Daniel Wujanz wrote: Thu Apr 29, 2021 8:35 amWhat Martin described is exactly the way we do it - we take the entire helix / kinematic point cloud / point sausage and cut it into "static" pieces. It is important to have no occlusions within a section. Interesting subject - I'll discuss it in greater detail in the PointCab webinar series in roughly a month.
Here was me calling them polycylinders, I reckon point sausages should definitely become the accepted standard term, certainly has my vote :D

The jargon surrounding all this is a bit counter-intuitive. e.g. are coordinates stored in an octree structured or unstructured? Coming from a survey background, I tend to think of raw, unadjusted, observed data (usually spherical polar, or cylindrical polar) and reduced, adjusted and corrected data (Cartesian or geodetic coordinates). Personally, I'm always happy to get the option of access to the raw data as this can provide far greater scope for algorithmic analysis, reduction and correction beyond the defaults provided by the equipment vendor. This is particularly true when combining images with scan data and getting rationalised points clouds as seen in NavVis.
Leica Pegasus2 (but not BLK2GO @Paul: Please note that this is not intended to be Leica-bashing - it is just not understandable from a client's point of view)
Also disappointed that trajectory and image matching data is not part of the standard BLK2GO output. It makes it an altogether less interesting piece of kit from the point of view of a 3rd party developer.
jedfrechette
V.I.P Member
V.I.P Member
Posts: 1039
Joined: Mon Jan 04, 2010 7:51 pm
Full Name: Jed Frechette
Company Details: Lidar Guys
Company Position Title: CEO and Lidar Supervisor
Country: USA
Linkedin Profile: Yes
Location: Albuquerque, NM
Has thanked: 28 times
Been thanked: 94 times
Contact:

Re: Point cloud formats, structured/unstructured

Post by jedfrechette »

smacl wrote: Wed May 05, 2021 4:43 pmThe jargon surrounding all this is a bit counter-intuitive. e.g. are coordinates stored in an octree structured or unstructured?
I was going to make that point when this thread started, but refrained thinking it was perhaps needlessly pedantic, so I'm glad somebody else did. :D

I think the somewhat vague and inconsistently used jargon is partially a result of the relative youth of this industry. Perhaps there should be a set of well defined processing levels for point clouds similar to what exists for remote sensing imagery data, e.g.

http://uregina.ca/piwowarj/Think/ProcessingLevels.html
Jed
Post Reply

Return to “General Chat”