What does the "Reduce the number of points of the 3D object" option do?

[Jimmy]

I asked this in another thread but I'm guessing that it got buried so I'll make a new post instead. I'm wondering what the option does more than "reduce the number..." since that doesn't really tell me that much. Does it implement a decimate, flatten or smooth algoritm to the shading analysis that's not visible on the object (because I can't see any changes)?

I'd like to know because if it does what I think it does it could affect my workflow. Also, does this option mean that PV*SOL can handle models that have more than 500000 vertices now? 

 

[developer_mh]

Hi Jimmy,
 
this function basically does what the description says. It reduces the number of points of the 3D objects in order to make the overall handling and especially the shading calculation more efficient. It is essentially an algorithm that detects similar triangles and melts them together. In this manner, the number of points is reduced. But since this reduction can lead to inaccuracies in the resulting planes, a warning is given that the user has to check carefully the shading results.
 
Hope that helps, kind regards,
Martin

 

[Jimmy]

Hi Martin,

 

As you describe it it sound like what's known in much of the 3D modelling world as a "decimate" function.

 

https://docs.blender.org/manual/en/latest/modeling/modifiers/generate/decimate.html

 

Regarding the second and third part of my question.

 

• How much does it decimate (reduce the number of triangles)? Is it a percentage (like 50% or 10%) or a fixed value (like down to 10000 triangles)?
• Does this allow models that are larger than 500000 triangles to be imported into PV*SOL?

I'm asking because I'm currently doing this (reducing the number of triangles) manually to my models before I import them into PV*SOL. And If I knew more precisely what the function does it could save me a lot of time since I wouldn't have to take those steps myself.

 

[developer_mh]

Hi Jimmy,

for your question I had to ask a colleague and this is what he replied:

Quote

We use a chain of DirectX 9 processes to simplify these incredibly complex drone models by first determining the vertices currently in the model and then comparing them to the maximum possible in PV*SOL.
We then determine how much the model needs to be simplified to get to the target size.

The simplification process takes place in 2 steps:

• Melting down the structures (DirectX Weld) - https://docs.microsoft.com/de-de/windows/win32/direct3d9/d3dxweldvertices
• Simplify the structures (DirectX Simplify) - https://docs.microsoft.com/de-de/windows/win32/direct3d9/d3dxsimplifymesh

--> The "Weld" with the setting "D3DXWELDEPSILONS_WELDPARTIALMATCHES" looks for identical corner points and replaces them with a common corner point (thus textures are lost or distorted), but the model keeps its structure. This allows the subsequent Simplify to work better.

--> The "Simplify" is a rather dull algorithm that replaces complex structures with simpler structures. It does not first replace approximately planar structures with surfaces, but "ears" or protruding structures are "melted off". This is how the DirectX-9 alogirithm works. It can happen that small or narrow shadowing objects on roofs or even relevant larger structures are halved in size or disappear. This must always be checked by the user.

The feature should therefore only be used if an object is only just too large for PV*SOL, because then the simplification measures should not have such an impact.
If the original number of points of the model exceeds the permissible number by a multiple, something really bizarre can result.
However, this makes the shadowing simulation much more performant. You always have to estimate it.

My tip: The user should first reduce all planar surfaces to the maximum in a professional 3D modelling tool, so that only a few corners are ever used there, he should delete all unnecessary ears and completely remove insignificant surrounding areas, then he can usually import it into PV*SOL. This is necessary if he doesn't want to spend hours simulating in PV*SOL afterwards, because we have to take every single triangle into account in the shading projection (above a certain size).

I hope this helps a little. It's one of the most complex topics in 3D and especially concerns these complex autogenerated meshes from drones, which always come out as a mush with millions of triangles.

 

I guess this answers both of your questions and perhaps sheds some light on the algorithms behind it.

Kind regards,

Martin

[Jimmy]

Thanks you so much Martin. I don't think that I will use the function as I feel more confident when I can check the mesh myself, I also get to keep all the "ears" that I want when I do it manually.

Still, I appreciate you making the effort to provide an answer for me.