Thank you so much for all the feedback! I'm sorry for the late reply, but i've been busy writing this massive post, hehe... so here we go:
Okay, here is a step by step guide on how to achieve this. It is long and a bit complicated, but bear with me. When you have done it a couple of times, it makes sense. There are basically 5 steps in achieving this:
- Capturing the data
- Getting it into Blender (free powerful 3d software)
- Combining and optimizing the captures in Blender.
- Getting the OBJ into sketchup:
- Working with the model in sketchup
The first two steps are largely covered in this great video that helped me alot in the beginning! I still urge you to read my description as well though, as some problems and tips are not covered in detail. But its a great starting point!:
https://www.youtube.com/watch?v=F_XsmoZJmG8
1. Capturing the data
Software used: RenderDoc, modified chrome browser
The first step is to capture the 3d data. This is done through a new software called RenderDoc, developed at MIT. The way it works is (very simplified) is by reading the actual data stream through the graphics card in a certain application. In this case, the browser.
This requires a modified shortcut to chrome, that allows you to intercept the data stream from the graphics card. We then inject this data stream into RenderDoc, so we can capture it. How all of this is done, is described in detail in the video linked above!
You then navigate to “earth.google.com” and use RenderDoc to capture the desired area(s). Google earth works with a number of different LOD’s (level of details), meaning that the further you zoom out, the more simplified the 3d mesh becomes, and the lower the texture resolution is.
With that in mind, it can be benefitial to do a number of captures with the view zoomed in, and then combining them later in blender. That way you can cover a larger area in higher fidelity.
Be aware that i’ve run into several problems with the actual captures and their readability in the next step. In my experience capture sizes of more than 120 mb will be hard to handle. If it is larger, try to zoom in a tad more, and the size should go down J. It is also very important that you really move the camera around during the actual seconds the capture happens! I use the “capture with delay” function set to 3 seconds, as this makes this much easier.
2. Getting it into Blender
Software used: Blender, blender plugin “Mapsmodelimporter”.
The next step is to get the captures into Blender. Blender is an extremely powerful 3d software, that is completely free. It has a number of very specific plugins that allows us to read and work with the captured data we just made. And the first of these is a plugin called “Mapsmodelimporter”. This allows us to read the .rdc files that RenderDoc created.
When you import the capture you will notize that it has alot of overlapping geometry, and that the capture probably looks a lot larger than you imagined. This is because it has saved all the LOD’s in the view, and you simply select all of the large planes containing the rough geometry you dont need, until you a left with just the detailed area you actually wanted to capture. It looks something like this:
3. Combining and optimizing the captures in Blender
Software used: Blender, Lily capture merger, Lily texture packer. Both can be bought cheaply here:
https://gumroad.com/eliemichel
In this step we will try to combine all our captures into a single, seamless model, and then clean it up and optimize it for exporting.
It is important to keep each capture in its own “collection” folder in blender, as this will make the merging much easier. we import one capture at a time and align it to the previous one, before moving on to the next.
We will use the plugin called Lily capture merger to align the two collections together. This video shows how it works:
https://www.youtube.com/watch?v=BcGM76dwF_s
when you are all done, you should have a number of collections in blender, and all the models should be perfectly aligned. something like this:
Notice all the 10 collections on the top right, each containing a single capture, representing a part of the complete model.
It is now time to erase overlapping geometry between the collections. This will remove flickering in enscape later on, and it will greatly reduce file sizes as well.
You do this by simply turning on one collection at a time, and deleting the squares that is present in another collection already. This is quite straight forward. When you are done, no part of any collection should overlap with a different one. you will notice that the captures are made up of smaller squares, and this makes it very easy to delete parts of it you dont need.
Now it is time to optimize the textures. When you make a capture from RenderDoc, all the textures are individual, resulting in more than a hundred different images for a single capture. We would like to reduce that to just one for each collection! For this, we use the plugin “Lily texture packer”
We create and save textures for one collection at a time, turning off all the collections we are not working with.
This video is a good introduction:
https://www.youtube.com/watch?v=OyaCfshX3zo&t=1s
remember to save the generated textures (go to the UV editor, find the newly packed image in the dropdown, and save it to a location).
Now it is time to link the new texture to the geometry. To do this, simply go back to the 3d viewport, select all the geometry in the collection, click “ctrl-L” and select “materials”.
Repeat this until all collections have ONE texture, all textures are saved, and all collections have linked the new textures to them.
Now we have to join all the individual sqares in each collection into one single mesh.
To do this, simply select all the geometry in each collection, and click “ctrl-J” (join). Now each collection should be selectable as one single mesh pr collection.
It is now time for some cleaning! RenderDoc creates a lot of overlapping geometry, and that makes the models very heavy, and hard to work with. It is fortunately very easy to fix in Blender.
Simply select each of the joined collections, click “Tab” (to go into edit mode), right click the mesh and select “merge verticies – By distance”. This merges all overlapping faces. In the bottom of the screen you should see blender tell you how many verticies it has removed. It is alot! But, for some reason it does not alway work in one go, so you have to repeat the command. If it says “0 verticies removed” next time, you are all good. If it still reports a number, continue to repeat until it says zero.
When this has been done for all collections you are done with blender. At this point you should have a number of collections (perfectly merged together), each containing one capture with one packed texture and cleaned geometry.
Now, simply select “file – export – obj” and save the file as an .obj 3d file.
and I agree completely. large changes must indeed be well considered, and thought through. And that is exactly why I'm pointing out that the current system is not working. It has to be looked at again for the software to be usable for serious work.
want to know if there is a tutorial