Jump to content

Using Google Earth to Generate a Reference 3D Model/Point Cloud in Vectorworks


Guest Alex Sagatov

Recommended Posts

Guest Alex Sagatov

Hello Forums! A post over on the wishlist inspired me to post this workflow I've been messing around with. A few people here have been using it in house for some projects recently!

 

I should also note that this workflow is by no means standardized and I wouldn’t trust elevation data or anything like that; however, from my testing so far, it seems to give decent enough results that I can plop a building in the results and it looks pretty nice.

 

1) Download Google Earth
2) Find an area you want to make in Vectorworks.
3) In short, pretend you are a drone. Spend a few minutes taking screenshots in Google Earth. Make sure to turn off labels and avoid getting any text in the screenshots. Make sure to get some different views and elevations to capture heights. Make sure to get some side shots, screenshots from different angles, different elevations, different distances, etc… but still focus on your area. Try not to get too much outside area in the screenshots.
4) Put all your screenshots into a folder.
5) Upload the folder to Vectorworks Cloud Services.
6) In Cloud Services, use the “Photos to 3D Model” command and give it a bit (In my experience, I am normally waiting about an hour)
7) Retrieve your point cloud and possibly a mesh! The Point Cloud will be ready first and tends to be much easier to work with. The mesh only works for me half the time in my testing so far, but it can give pretty cool results!

 

Example results!

 

150 Screenshots of a town near our office - You will notice the mesh actually got a bridge properly!

 

Screen Shot 2019-02-01 at 12.52.56 PM (1).png

 

55 Screenshots of part of the Grand Canyon!

 

Grand Canyon (2).png

 

Link to comment

WOW...what is needed is a trim command...but here is the thing...when you trim it does not delete but just gets sent to another layer that is turned off...this way you can cut holes to show various options or timing as in "This is the existing conditions"...."This is phase 1"...."This is phase 2" etc

 

the trim tool needs to be dynamic in that if i adjust the clip poly then the edges of the earth update also on all layers  & you still have only one DTM

 

Needs to have a fill depth...

 

 

Edited by digitalcarbon
Link to comment
Guest Alex Sagatov

@digitalcarbon  Wow! That is actually a pretty great result for only 36 images! I haven't personally tested with less than 50.

 

For what its worth, I have found the point cloud much easier to work with. The mesh may look better (especially from a distance), but the meshes tend to be extremely complex. With either option, I would make sure to throw it onto its own design layer so you can easily toggle it on/off as needed.

 

The resulting Point Cloud/Mesh are not georeferenced in any way; however, you could overlap it with other information that may be in a georeferenced layer which should then at least get it close to the UTM grid. (I am not familiar with this enough to run a full test. Maybe someone else can?)

 

In my tests so far, I have found the best methods of clipping are using the Point Cloud and then isolating the point cloud as needed. In a few cases, I ended up with multiple copies of the Point Cloud in order to crop it multiple different ways.

 

My test with the Grand Canyon also brought me to the idea of trying to create a site model out of the result. At least in my tests, the point clouds were millions of points. Do not do try that. haha.

 

Here is a workflow I am messing with currently:

 

Use Isolate Points on the point cloud to crop the area in question. Delete the other points.

 

Screen Shot 2019-05-02 at 10.31.04 AM.png


In your workspace editor, add the menu command for Extract 3D Loci from Point Cloud...

 

Screen Shot 2019-05-02 at 10.34.17 AM.png


Use that to extract ~30,000 or so points from the remaining point cloud. Use those to create a site model...

 

Screen Shot 2019-05-02 at 10.52.26 AM.png


Again, I wouldn't trust these elevations if you are doing site analysis or anything where you need specific survey data... but as a quick way of getting  a cool site model or something to generally work with, it can work. It would require some clean up though as you can see my example.

Link to comment

what is needed is to start with the general point cloud to DTM as a starting point for a project ..

 

then you mark out what needs to be surveyed on the rough DTM

 

then when you get the quality topo from surveyor (which would only be a portion) you clip out a hole in the rough DTM and add in the quality DTM...but still keep the rough as context DTM...understand?

Link to comment
Guest Alex Sagatov

@digitalcarbon I will mess with this a bit more in the coming days and try a more real world example, but off the top of my head and to make sure I am understanding:

 

In my Grand Canyon example, I created a generic site using the method I outlined in the above post. In a 3D view, I set my site model to 3D mesh and ungrouped it. This gave me a mesh object for the top of the site. I could then theoretically take my property line, trace it with a polygon, and extrude that polygon. I could then extrude the polygon and subtract solids from the mesh object. This would then leave a hole for me to put in my real survey data:

 

Screen Shot 2019-05-02 at 11.24.44 AM.png

Link to comment
Guest Alex Sagatov

For what its worth, I know we are working on a more developed and clean workflow/tools for bringing in georeferenced images and topography. I will keep my fingers crossed that we see this in a close version! 😀

 

 

Hopefully this workflow can still be useful in someway though!

Link to comment
17 hours ago, digitalcarbon said:

also need a 3 point flatness tool...if i know the elevations of 3 points then i can set those points with plumb bobs  then pick those 3 points on the point cloud and say align to plumb bobs...this will set the proper orientation...

 

There is a tool that I use for quickly flattening variously rotated extrudes to the same working plane. First (if necessary) set your working plane to be flat to X,Y.

 

Then select the object, and choose "Align Plane" (behind the set working plane, in the 3d tools)

image.thumb.png.7f8398f1f6e8321aabb762b18b2354aa.png

 

Choose 3 points on that object that you would like to be co-planar (A bit like setting a working plane, but this works in reverse) and boom, on the 3rd click, the model will "un-rotate" to flat.

 

Not sure if this works for point clouds, but it's a massive time saver for general geometry alignment.

 

  • Like 1
Link to comment
  • 8 months later...
On 5/1/2019 at 3:40 PM, Alex Sagatov said:

3) In short, pretend you are a drone. Spend a few minutes taking screenshots in Google Earth. Make sure to turn off labels and avoid getting any text in the screenshots. Make sure to get some different views and elevations to capture heights. Make sure to get some side shots, screenshots from different angles, different elevations, different distances, etc… but still focus on your area. Try not to get too much outside area in the screenshots.
 

 

 

@Alex Sagatov  How do you make your screen shots?  What function or app do you use to make these?

Link to comment
Guest Alex Sagatov
12 minutes ago, Markvl said:

@Alex Sagatov  How do you make your screen shots?  What function or app do you use to make these?

 

I personally did these on Mac. On Mac, I was using Command + Shift + 4 to take quick screenshots that saved directly to my desktop. This made the process pretty fast. I did do this once on a PC where I used the Snipping Tool. IIRC, there is a shortcut for quick screenshots using the snipping tool, but I don't recall the shortcut off the top of my head.

Link to comment

Photogrammetry needs a lot of CPU power. Some programs also run with CUDA, so they use the graphics card instead of the CPU.
For a good point cloud you want to use several hundred photos. The calculation time increases overproportionally with the number of photos and takes about 15 minutes for 300 photos, one hour for 600, 3 hours for one thousand and two days for 4000 photos. (using VisualSFM with CUDA on a RTX2070)

 

VisualSFM is OpenSource if you want to try out.

Edited by herbieherb
  • Like 1
Link to comment

Thanks @herbieherb for your very informative response.

 

I see VisualFSM has a Mac version, so I’ll give it a try. How do you bring your end product from VisualFSM into vectoworks?

 

Also, @herbieherb- you seem clued on about GPU, so if I may ask you a few questions, I can’t find any definitive answers on

these:

 

(1) Would you know how the RTX 2070 compare to Quadro M or P series. I’m in particular interested in performance for VisualFSM, Vectorworks, or GPU renderer like Lumion/Twinmotion.

 

(2) Some call Quadro ‘workstation’ GPU and RTX and the like ‘gaming’ GPU. Price is drastically different, how is this difference in price justified?

 

(3) Do you think a PCIE 3.0 GPU like Quadro M6000 on a PCIE 2.0 x16 slot will greatly hinder its performance? Some forum posts says its almost negligent at others say its wasteful. 


Apologies to all to turn this topic so geeky. I’m in the processing of choosing which upgrade path to take for our Vectorworks-led BIM workflow.

Link to comment

Quadro graphics cards only give you more performance in the software for which special drivers are available. These custom drivers are one factor that makes Quadro so expensive.

The Quadro cards have the same graphics chips as the Geforce cards, are simply technically throttled, so that reliability and lifetime are higher. A Quadro RTX 4000 has the chip of the Geforce RTX 2080 Ti, but it is throttled to the speed of the RTX 2070 Super. They have also built in more RAM. That is the other reason why they seem so expensive for the performance they provide.

But these special drivers and specs makes them more suitable for CAD which are specialized in this. (calculate a lot of geometry without fancy shaders) but slower for gaming. (Less geometry, but with fancy shaders). Lumion and Twinmotion run with a game engine and therefore the Geforce cards are more suitable. Also Vectorworks runs better with the gaming drivers of the Geforce, because it is optimized for it.

For Vectorworks, Lumion and Twinmotion a Geforce is therefore the better choice.

For further technical advice, post the parts of your current computer in the hardware section of the forum.

Edited by herbieherb
  • Like 2
Link to comment
  • 3 years later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...