Mesh workspace: reduce or remesh 3D scan to keep the Model precise

[RaceCarEngineer]

Most 3d scans have far too many faces, so Fusion 360 is slow down a lot. 

What method of reducing face count is the best to keep geometry as close to the original? Especially holes, threads and edges should be kept as precise as possible to the original for reverse engineering purposes. 

Options I found are: 
- use remesh with a factor 

• adaptive
• uniform

- use reduce 

• Tolerance
• Adaptiv
• uniform
• fixed amount of faces 

Which method does the best job to get for example from 6million to 300k faces?

[laughingcreek]

what are you hoping to do with the mesh?  being a scanned mesh, you are very unlikely to get anything useful from any type of workflow that involves converting the mesh to a brep directly. (well, there are some organic shapes that covert well, but sounds like your talking about a mechanical type object).  reducing the mesh in any way is just going to make these conversions even worse.

the most common workflow is to use the mesh as a visual reference only, maybe extracting some mesh cross sections, and rebuilding the object natively from scratch.  if you you do it this way, then reducing the mesh doesn't do anything for you.

short answer-don't reduce the mesh. don't convert the mesh to a brep directly.

[aliobidi]

best method import the file into Blender  .. watch YouTube 

[TrippyLighting]

I suggest you share at least a screenshot.

A Mesh per definition is NOT precise. As opposed to CAD geometry it has a fixed resolution.

@aliobidi there is now a built-in version within the Product Design Extension. I have yet to try that one out. There is also a Fusion 360 addin available that does a very nice job (Don't have the link handy)

I've achieved very good results using InstantMeshes.

Of course all of that makes only sense for organic meshes.

[aliobidi]

interesting .. soon I'll deep in API 

[RaceCarEngineer]

First I need the parts to reverse engineer bolt pattern, mounting points etc. This means the holes, diameter or in general dimension must be retained as dimensionally accurate as possible. The problem is that if you import models over about 100 MB into fusion360, it gets quite slow. if the mesh is reduced, holes, diameter etc. get less accurate. 
I can reduce the mesh size already on the 3D scan software, however, I found I have more accurate results reducing it only in fusion. Here I can select only non-critical parts of the model to reduce the mesh. However often the model is still too big and I need also to reduce the critical parts. 
So long story short, as initial requested which of the offered method does the best job, with keeping accurate dimensions. 

BTW, in the meantime, I've done some testing and found remeshing doesn't always reduce mesh size, even if both sliders are all to the right. Reducing mesh with Tolerance several times keeps the model nice, while uniform reduction gave the worst results. Just want to understand more what fusion exactly does in the background. 

[RaceCarEngineer]

Here is a comparison.

- left is heavily reduced mesh in Fusion 360 with 67k facettes. I made face groups for the holes and reduced only the mesh around it to the point fusion could only reduce the mesh more when the whole part was reduced. 

- right is full-resolution STL with 0.1mm meshing straight out of scanner software

I think this picture illustrates the problem quite well. With 67k facets, the model is still quite big for fusion360, but the holes and geometry are already heavily influenced. This is only a car alternator. If I going to model a whole engine compartment I will have x million facets in Fusion, which slows it done a lot. 
So I need to find the best methods to keep the geometry intact while reducing facets count as much as possible. 

[TheCADWhisperer]

---

@RaceCarEngineer wrote:

… as precise as possible to the original for reverse engineering purposes. 

---

Use the mesh only as reference along with precision measuring instruments and remodel the part from scratch. Period.

[RaceCarEngineer]

Sorry, this answer is not a solution to my question. Of course, some critical measurements can easily be verified with the usual hand measuring tools. However, with the 3D scanner reaching a precision of 0.01 mm, there is no need to measure everything again with a Faro arm. Some basic manual verification measurements are enough for brackets and parts with non-tight tolerance. And then I want to have the model as precise as possible in Fusion. At least the connection points. In the example of the alternator, connection points are different in all XYZ coordinates. BTW Autodesk has made a few webinars about that topic. So it is not that I'm the only one needing 3D scans in Fusion360...

[jeff_strater]

"I need to find the best methods to keep the geometry intact while reducing facets count as much as possible"

I think you are starting to realize that these are competing goals.  Reducing facets, by definition, reduces accuracy.  You can get part way there by doing selective Reduce or Remesh in mesh mode.  If you want to stay in parametric mesh mode, first generate face groups.  Otherwise, use the Direct Edit mode to select individual facets, and Reduce or Remesh those.  Either of these, though, will be quite tedious.  But, you can partly achieve your goal of selective simplification this way.  No guarantee that you can get all the way there, though.

I think it is pretty safe to say that Fusion is not designed for dealing with scan mesh data.  There are probably other tools that are more oriented towards this kind of data, though I don't know what they are.  Fusion is really aimed at reverse engineering for mesh data:  Use the mesh as a reference to re-create more native geometry. 

I don't think you ever answered @laughingcreek's question:  What is your end goal with this data?  Say you were able to get to a reasonable number of facets, and with the accuracy that you want.  What next?  What do you want to do in Fusion with this data?  Do you plan to try to convert this to a BRep, and do solid or surface modeling on it?  What edits do you want to make to the scan?  Will you try to 3D print it, machine it, or use it for reference to build other parts?  That information will help us determine where to send you next....

[laughingcreek]

There isn't a way to reduce a mesh AND keep it precise.  anytime you do ANY kind of edit or conversion to a mesh you will lose something.  there is no way around that.

and there won't be a single reduction method that is a best fit for all possibilities.  you might try using a mesh specific software (like mesh mixer) to isolate parts of the design and break it into smaller more manageable pieces.  then pull in the smaller pieces and work on them that way.   You can do it in such a way that the pieces will com into fusion located in the right place relative to each other.

[RaceCarEngineer]

Thank you for the answers. 

I understand quite well that Fusion is not the best software to deal with scan data. Because of that, I  prepare the models as good as possible in the scanner software. Most of the time I need the scan data as a reference to design new parts, which bolts onto them. For example, I scan an engine without side aggregates fitted and the engine compartment. So my task is to design a new engine, alternator brackets, and parts that bolt on the engine and must fit into the engine compartment. So I need the whole engine compartment and engine model in Fusion, to be able to properly place components (alternator, steering pump, steering rack, design new water pump and much more). 

Back to my initial question. I already tried the different reduction and remeshing methods in Fusion and found with the same reduced face counts, the part comes out differently and some methods keep the geometric topology better than other. It's absolutely clear to me, that with any face conversation and especially reducing faces details are lost. However, a smart algorithm makes an adaptive mesh that keeps edges, and holes high resolution and reduces it a lot on flat surfaces, constant radius etc.  Other methods may have the target, to make the mesh look as nice as possible or distribute the mesh triangles as smooth as possible, rather than keep the geometry as close as possible. 
So my specific question is which of the methods offered in Fusion 360 prioritises geometry most? 

[TrippyLighting]

There is no single method better than any other. That is exactly the problem with scanned data or in general triangulated meshes. They don't have any topology in a way a real CAD model or a quad mesh have. 

So the work that discerns between areas that need to be very accurate, e.g. bolt and mounting locations, and  geometry that is needed mostly for visual reference and to judge available space, is left to the user.

I would section/split the mesh into those parts that need to be accurate and those that can be strongly reduced and then use different methods on those different mesh sections.