Jump to content
Andy Broomell

Photogrammetry is Coming

Recommended Posts

IMG_1918.PNG.2edb2144e1dc843d5009861d805b4e22.PNG

 

 

Hadn't heard about this until the above Instagram post. Was this announced at the Summit? 

 

Sounds cool! What hardware is needed for such things? I have no experience in photogrammetry / lidar.

  • Like 1

Share this post


Link to post

Although zoomer is right that I do share similar stuff, this new tech is actually out of my league! The method shown above is actually from taking a large set of photos of an object from many different angles and then compositing a mesh or point cloud out of the set.

 

All I do is take a 2D image and give it height based on the brightness of the pixels :) 

Share this post


Link to post

I did this with a ribbon frieze on a wall with a bump shader once. Worked well. Picture a bit grainy, but you get the idea!

 

 

Picture1.png

Edited by markdd

Share this post


Link to post

Ooops,

should learn to read.
Just looked at the image and it looked familiar :)

 

But something like this was asked in another thread ....

Edited by zoomer

Share this post


Link to post

Some of it is done with expensive, complex laser scanners, but one of our R&D tasks is to see how we can get "scans" of both interior and exterior geometry derived from multiple smartphone pictures. Its REALLY COOL.

Share this post


Link to post

So I could take a bunch of photos of a particular couch in a furniture warehouse, then take it into Vectorworks and have a 3D model of that exact couch in the set design rendering? That would be amazing!

  • Like 1

Share this post


Link to post
3 minutes ago, Andy Broomell said:

So I could take a bunch of photos of a particular couch in a furniture warehouse, then take it into Vectorworks and have a 3D model of that exact couch in the set design rendering? That would be amazing!


That's the idea!

  • Like 1

Share this post


Link to post

As long as they aren't left as extremely overcomplex mesh objects, they seem to be doing quite well. I've seen the mesh simplification from 2018 applied to them and they keep their visual quality while being reduced to a workable poly count.

However, this is just yet ANOTHER reason I want a real and reliable way of converting Mesh geometry, with textures, into a Vectorworks Solid. Preferably on the automated side.

  • Like 2

Share this post


Link to post

It sound like this is the new "hype" in CAD/GIS/3D packages, as a GIS progam that I use is working on introducing basically the same thing but then from UAV aerial photos.

Would this work for aerial photos as well in Vectorworks? That would be really nice for Landscape users. Pls. Pls. Pls. :)

Share this post


Link to post
12 minutes ago, Art V said:

Would this work for aerial photos as well in Vectorworks? That would be really nice for Landscape users. Pls. Pls. Pls. :)

Possibly, however a lot of the information used to compile the 3D surface from 2D images is derived from both GPS coordinates AND the accelerometer on the phone, so it sort of knows where it was when it took a single picture in relation to a number of other single pictures. I am fairly certain this extra meta data attached to the images is needed to compile a mesh and that you cant use the same process retroactively on simpler 2D images.

So i think it might be possible to newly take aerial photos and create a site from them, but not to create a site from existing images.

Share this post


Link to post

So it looks like we could start looking into UAV/drone equipment for aerial photos in the near future. It starts to look like Agisoft's Photoscan is getting some competition on multiple fronts in this area for those with relatively simple/straightforward needs.

Share this post


Link to post

I did some limited experimentations with photogrammetry via an entry level DSLR a year or two ago using specialized software that I convinced my school to get me a license for.  It's not a particularly straightforward process, nor did it generate particularly useful geometry at that point.  There's no reason at that stage I couldn't have done it with smartphone photos.  Basically every camera on earth at this point has sufficient resolution and sensor size for photogrammetry at this point.

 

The issue is (as others have said) getting it to useful geometry.  Photogrammetry is TOO accurate in my experience so far for easy architectural use/and it can't distinguish between objects.  And working with point clouds is somewhat difficult.  I was attempting to use it for site surveying a historical site in Malta where I had limited time on site and no ability to return after an initial visit.  I ended up doing the surveying I needed manually and using the photogrammetry as a reference.  Fortunately that was a school project so it didn't matter when I fudged things to deal with inconsistencies.

 

I know some archaeologists have been using photogrammetry for documenting excavated objects for a while (which was my initial introduction to it).  They seemed to keep the resulting models as records that lived entirely in the specialized photogrammetry programs they used - I have no idea if/when they were extracting that data to another modeling program.  I did not have a chance to probe that too much, as I was only onsite with those guys for a few days on a somewhat remote site in Peru.

Share this post


Link to post
57 minutes ago, JimW said:

Possibly, however a lot of the information used to compile the 3D surface from 2D images is derived from both GPS coordinates AND the accelerometer on the phone, so it sort of knows where it was when it took a single picture in relation to a number of other single pictures. I am fairly certain this extra meta data attached to the images is needed to compile a mesh and that you cant use the same process retroactively on simpler 2D images.

So i think it might be possible to newly take aerial photos and create a site from them, but not to create a site from existing images.

 

 

You shouldn't need any metadata for photogrammetry - all the info is based on the photos, you just need enough photos from sufficient angles (at least with the applications I have experience with)

Share this post


Link to post
1 hour ago, Christiaan said:

How efficient will these objects be? Sounds like they'd be very data heavy.

Tricky to answer.

You generally need to optimise the mesh a bit, meshlabs and 3ds max are good for this.

Also the UVW maps can be absolutely bonkers, unless you do an unfold operation before

This was done with a 50mpix canon last year

 

But also.... Capturing reality has a great free trial version you can play with. Obv, the higher the image quality, the better the finished article. The vid above was done in Agisoft Photoscan.

https://www.capturingreality.com/

HTH

  • Like 1

Share this post


Link to post
1 hour ago, Art V said:

Would this work for aerial photos as well in Vectorworks?

Since they brought point cloud support in, yes, it works extremely well.

I did a talk through for some friends last year; but I do babble on (ignore the first 5 minutes or so)

https://www.youtube.com/watch?v=9E4N2u_2RWU&t=555s

And an example point cloud is here;

https://www.youtube.com/watch?v=O4LbOLUkOgk

My friends farm is about a mile long, and half a mile wide, and was scanned in 3 flights, over 45 minutes, and the whole vally consists of around 60 million points

Be very careful about operating inside the law though. UAV operators are getting some bad press at the moment.

 

Edited by RussU
  • Like 1

Share this post


Link to post
41 minutes ago, nrkuhl said:

 

 

You shouldn't need any metadata for photogrammetry - all the info is based on the photos, you just need enough photos from sufficient angles (at least with the applications I have experience with)


Apparently there are a number of methods, one just uses ONLY 2d imagery and another type is based on recording where the device/phone was in 3D space when it took a 2d image to reconstruct a photo. I do not know specifically which or in what combination our demo was done in, but they plan to roll it into something like Nomad and told me that loading in older photos may not work at all if they weren't taken with the app. I don't pretend to understand it fully, but apparently its way easier to get a 3d model at the end with fewer source images if you include positional data.

  • Like 1

Share this post


Link to post
1 hour ago, JimW said:


Apparently there are a number of methods, one just uses ONLY 2d imagery and another type is based on recording where the device/phone was in 3D space when it took a 2d image to reconstruct a photo. I do not know specifically which or in what combination our demo was done in, but they plan to roll it into something like Nomad and told me that loading in older photos may not work at all if they weren't taken with the app. I don't pretend to understand it fully, but apparently its way easier to get a 3d model at the end with fewer source images if you include positional data.

Does this mean that if you want to use photos they have to be made with the Nomad app to make it work within Vectorworks? I.e. no other sources at all?

Share this post


Link to post
5 minutes ago, Art V said:

Does this mean that if you want to use photos they have to be made with the Nomad app to make it work within Vectorworks? I.e. no other sources at all?


For all types of photogrammetry this is not the case, but in the limited tests I saw it was. However this is a discussion of a feature that is early in R&D and any answer I give you now could easily become invalid before its maturation.

Share this post


Link to post
3 hours ago, Art V said:

It sound like this is the new "hype" in CAD/GIS/3D packages, as a GIS progam that I use is working on introducing basically the same thing but then from UAV aerial photos.

Would this work for aerial photos as well in Vectorworks? That would be really nice for Landscape users. Pls. Pls. Pls. :)

Check out this thread for more on alternatives for using aerial photos and satellite images to create meshes:

 

https://forum.vectorworks.net/index.php?/topic/51710-working-with-lidar/

Share this post


Link to post
13 hours ago, JimW said:


For all types of photogrammetry this is not the case, but in the limited tests I saw it was. However this is a discussion of a feature that is early in R&D and any answer I give you now could easily become invalid before its maturation.

Thanks for the clarification. Let's hope it will be possible to use 3rd party sources, e.g. when a client provides the images made by a UAV, as its usability would otherwise  be too limited for Landscape users.

Share this post


Link to post
13 hours ago, jmanganelli said:

Check out this thread for more on alternatives for using aerial photos and satellite images to create meshes:

 

https://forum.vectorworks.net/index.php?/topic/51710-working-with-lidar/

Thanks for the link, I did already read that one and saved the link for the references. What I meant to say was that a few days ago the GIS software company announced they were looking for beta testers for a photogrammetry tool inside the GIS software and now there is discussion of Vectorworks (also) having a photogrammetry  tool as part of its toolset. Hence the "hype" comment as I expect more CAD/GIS/3D software developers are going to implement this as part of the software instead of relying on 3rd party tools. Though the 3rd party tools that are specific to photogrammetry will prolbably still have more advanced functionality/options for a while (.e.g. Agisoft Photoscan Professional)

For me the interesting part is that I can use Vectorworks for basic (or even more than basic ?) conversion from photo to 3D within the software and have native objects instead of having to get/use another tool which may introduce another format conversion on top of the photo to 3D conversion.

Share this post


Link to post

Just to follow up on this, this is the first publicly released feature of ours related to this topic:
 

 

On 9/28/2017 at 3:41 AM, Art V said:

Thanks for the clarification. Let's hope it will be possible to use 3rd party sources, e.g. when a client provides the images made by a UAV, as its usability would otherwise  be too limited for Landscape users.

 

This is the case! images from any source can be used.
 

  • Like 3

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

 

7150 Riverwood Drive, Columbia, Maryland 21046, USA   |   Contact Us:   410-290-5114

 

© 2018 Vectorworks, Inc. All Rights Reserved. Vectorworks, Inc. is part of the Nemetschek Group.

×