A recent example of using mesh vs. point cloud
The new Apple iPad Pro LiDAR is equipped with a rear 3-D camera generating point cloud at the rate of about 500 points in 60 Hz, i.e. max 30,000 pts/sec.
Late March 2020, Apple had not provided developers with the real-time access API to the point cloud stream, but for the delaying meshes generated from the point cloud stream.
Now, late June 2020, Apple provides the real-time access API to the point cloud stream.
Mesh and point cloud have inherently advantages and disadvantages.
From the viewpoint of signal processing:
- Point cloud is the raw information
- Meshes destroy the raw information irreversibly.
From the viewpoint of presentation:
- Point cloud is hard to get in grip
- Meshes are visually comfortable,
but!, only provided with surface textures.
Signal processing is done by machine.
Presentation is for Human.
As data processing expert, we prefer absolutely the point cloud to the meshes!
For business, meshes with textures may be preferable to point-cloud with intensities.
Clients will decide.
Why deliver meshes
-
- V.I.P Member
- Posts: 627
- Joined: Sun Jul 24, 2016 6:19 am
- 7
- Full Name: Ioannis Papas
- Company Details: techical office
- Company Position Title: CEO
- Country: EU
- Skype Name: -
- Linkedin Profile: Yes
- Has thanked: 99 times
- Been thanked: 62 times
Re: Why deliver meshes
sim.herrod wrote: ↑Thu Jun 25, 2020 4:11 pm As with the 'why colourise point clouds' question that was asked a while ago, the most simple answer is...because clients ask and pay for it.
We use 3DR to create meshes, and once you've had a play with what settings work, a nice mesh is only a couple of button clicks away.
As has also been said, not that many of our clients have the ability to look at point clouds, but most use some sort of CAD package that can use mesh formats.
Can you please upload some examples from 3dr? I have heard that is a very powerful program and easy to use.
-
- V.I.P Member
- Posts: 316
- Joined: Thu Oct 15, 2009 12:51 pm
- 14
- Full Name: Martin Ankor
- Company Details: Anon
- Company Position Title: Surveyor
- Country: Australia
- Has thanked: 6 times
- Been thanked: 24 times
Re: Why deliver meshes
Sometimes.landmeterbeuckx wrote: ↑Thu Jun 25, 2020 10:38 am So if you have a nice cleaned pointcloud a mesh isn't that hard to do?
I should find some time to investigate Cloudcompare a bit more, not only for meshing.
A nice, clean object or structure can result in a very good mesh.
A complex structure, with lattice work, vegetation, etc can be an absolute horror to get a good mesh done, especially if aiming for something 'waterproof'.
This is one of the reasons why I like Cloud Compare's meshing method. The Poisson reconstruction is very good at avoiding reversed and conflicting normals or holes in the mesh. Having said that, even CC's method will have problems if you have two surfaces close together, with large enough point spacing for the mesh to bounce from one surface to the other.
- landmeterbeuckx
- V.I.P Member
- Posts: 1615
- Joined: Tue May 01, 2012 5:19 pm
- 11
- Full Name: Lieven Beuckx
- Company Details: Studiebureau Beuckx
- Company Position Title: Owner
- Country: Belgium
- Linkedin Profile: Yes
- Has thanked: 183 times
- Been thanked: 548 times
Re: Why deliver meshes
That's a pretty solid statement. I'll have to look a bit further into this. ThanksJoon wrote: ↑Thu Jun 25, 2020 6:46 pm A recent example of using mesh vs. point cloud
The new Apple iPad Pro LiDAR is equipped with a rear 3-D camera generating point cloud at the rate of about 500 points in 60 Hz, i.e. max 30,000 pts/sec.
Late March 2020, Apple had not provided developers with the real-time access API to the point cloud stream, but for the delaying meshes generated from the point cloud stream.
Now, late June 2020, Apple provides the real-time access API to the point cloud stream.
Mesh and point cloud have inherently advantages and disadvantages.
From the viewpoint of signal processing:
- Point cloud is the raw information
- Meshes destroy the raw information irreversibly.
From the viewpoint of presentation:
- Point cloud is hard to get in grip
- Meshes are visually comfortable,
but!, only provided with surface textures.
Signal processing is done by machine.
Presentation is for Human.
As data processing expert, we prefer absolutely the point cloud to the meshes!
For business, meshes with textures may be preferable to point-cloud with intensities.
Clients will decide.
- smacl
- Global Moderator
- Posts: 1409
- Joined: Tue Jan 25, 2011 5:12 pm
- 13
- Full Name: Shane MacLaughlin
- Company Details: Atlas Computers Ltd
- Company Position Title: Managing Director
- Country: Ireland
- Linkedin Profile: Yes
- Location: Ireland
- Has thanked: 627 times
- Been thanked: 657 times
- Contact:
Re: Why deliver meshes
I think clients will opt for what provides them with most value for their budget, if you can add extra value you can get sometimes extend that budget. Looking at the work that is being done with point cloud analysis, including your own, I think this is primarily being driven by developers looking to create that chargeable value. I have no doubt that we'll have increasingly robust object identification and abstraction from point clouds going forward, starting with simple geometric primitives, moving on to bounded primitives and then compound primitives representing known objects. Meshes are a useful component here for modeling deformities to an extracted primitive object, e.g. a dent in a sphere or a crack in a wall. Some objects are so irregular that adaptive meshes are the most cost effective way to represent them, e.g. the stone wall of a an old church. From a user value point of view this is possibly best handled through level of detail (LOD) arguments passed when accessing the data. The original data includes scans, photography, extracted geometric objects and meshes and the user can extract a subset based on the LOD needed for their intended use. We'll also have multiple copies of all this to examine change over time. Meshes are just part of the jigsaw and in this code monkeys humble opinion we've a very long way still to go to finish it.Joon wrote: ↑Thu Jun 25, 2020 6:46 pm A recent example of using mesh vs. point cloud
The new Apple iPad Pro LiDAR is equipped with a rear 3-D camera generating point cloud at the rate of about 500 points in 60 Hz, i.e. max 30,000 pts/sec.
Late March 2020, Apple had not provided developers with the real-time access API to the point cloud stream, but for the delaying meshes generated from the point cloud stream.
Now, late June 2020, Apple provides the real-time access API to the point cloud stream.
Mesh and point cloud have inherently advantages and disadvantages.
From the viewpoint of signal processing:
- Point cloud is the raw information
- Meshes destroy the raw information irreversibly.
From the viewpoint of presentation:
- Point cloud is hard to get in grip
- Meshes are visually comfortable,
but!, only provided with surface textures.
Signal processing is done by machine.
Presentation is for Human.
As data processing expert, we prefer absolutely the point cloud to the meshes!
For business, meshes with textures may be preferable to point-cloud with intensities.
Clients will decide.