how to model a lobster antennule? (n00b question)

Hi all, I am a PhD student in engineering, and a complete n00b to Blender and CAD software (though I’ve been reading…). I need to create a digital model of a lobster antennule (long story there…) that can be fed into a fluid dynamics code that solves for water flow around the (rigid) structure. I would be going off of actual measurements of real specimens - the model must be completely well defined (no ambiguous lengths, angles, or features that are not known precisely). I’m not sure that Blender is the right program for this, and if not, I am open to commercial software (there’s a fair chance I can obtain even high-end software like Pro/E or NX since I’m an academic user at a large university). Any advice on what program is best, or good modeling techniques for this problem, is very much appreciated! Here are my concerns/needs:

Since a lobster antennule is somewhat “curvy,” perhaps a combination of constructive geometry and free form modeling might be best? The structure consists of a cylindrical stalk, with many rows of hairs (some straight, some curved) lining the stalk in a specific, repeating pattern and orientation. The physical scales span centimeters (stalk length) to micrometers (hair diameter), though initially I may just do a millimeter long subsection. I’d like the ability to easily modify model properties (change an orientation angle or hair diameter or number of hairs).

The export format would ideally be a text file consisting of all the vertex coordinates of a surface mesh. All I need to model and save is the surface shape. It sounds like the STL format may work. The need for a universal, simple export format is critical for use in the CFD code later.

Another critical need: I need to be able to output a very high resolution mesh (ideally, an arbitrary number of surface coordinates) to the export file. I’d like to be able to vary the output mesh resolution while maintaining a “perfect” (read: mathematical, continuous) description of the model internally. Is this even possible? It sounds like spline-based NURBS may be better than a mesh model here, since a mesh model is locked into a given resolution?

Would Blender work for this? Or would something else be better, ignoring factors such as cost? A steep learning curve is OK, but I’d rather not learn a program only to discover it won’t work for this. Thanks for any advice (and sorry about the long n00b post).

I’m probably not the best person to answer your questions, considering that I’m pretty new to Blender myself. But after reading your post, there’s one thing that you may have trouble with. The hair part. Blender does have a build in particle engine, and it does support hair. However, I’m not sure how it exports, or if it actually occupies any actual space.

The other option would be to model the hairs. This wouldn’t be particularly difficult, I suppose, although size, and getting it attached to the rest of the mesh may be pretty hard.

But don’t take everything I say as final. Like I said, I’m pretty new to this.

AS far as modeling goes, there is nothing blender can’t do, and do very well,
no details to fine and no measurments to small for blender,
And you can export as a text file,
I think the only hang up you will have is learning everything you need to know and achiving the level of detail you are after, which is only a limmitation of you, not the software,
How long do you have to do this?

A good thing to look into would be “dupli verts” to place an object (hair) on every vertex of another object (lobster claw)
as well as “dupli groups” to recreate the hairs without recreating the verts and in turn the computaion load of more objects, you may neew a few different dupli groups for different length hairs, yes you will need to modle most of the hair, well at least a few strands and then dupligroups, and if particles do export as text, you can use them to make a nice fuzzy clay, if that is needed for acuracy.

I never knew blender was resolution dependent, so I dont know about “mesh resolution”
hope this helps

I hope this isnt the “lobster claw” you want to modle?


Thanks for the comments! I should have attached the following pics of a physical model first. I need to draw something similar to this physical model. It’s not that complicated looking, right?

Sorry I wasn’t too clear about the hairs. Indeed, I need to model each individual hair as a separate object attached to the stalk. And the entire thing is actually rigid (at least for my purposes), so the term “hair” may not be the best description…

How much time is a good question… I’d like to have a very basic model of a stalk and a few straight hairs within a few months to appease my adviser. The full fledged model might not be necessary until this summer or later.

Again, I should be clearer about those hairs - I don’t actually have thousands of hairs covering the stalk, it is actually a precise zig-zag pattern of individual hairs (maybe 20 per repeating row) along with some curved hairs. See attached pics. So unlike hairs on a human head, these hairs are rigid, precisely oriented, and need to be individually defined (not fuzzy). I agree that some kind of “duplicate” tool would be great - thanks for the info.

Being a n00b, I’m not sure how exactly a mesh model works, but I thought that the surface is defined by a large number of vertices that make up a triangular mesh? So, the resolution (by resolution I mean smoothness of the surface) is essentially defined by the number of vertices, or equivalently, the size of each triangle, right? Ideally, I’d like to specify the resolution of the mesh I export, but retain a smooth, continuous (spline based?) surface description in the modeling program, since the actual surface is in reality completely smooth and not “piecewise triangular.” Another option is to simply have a single, extremely high resolution mesh that will always be sufficiently more refined than my CFD code’s grid, however, it is very difficult to predict what that resolution should be, so I’d rather the first idea. Am I understanding this stuff correctly?

Thankfully, I do not wish to model the claw in that pic… Actually, I am not modeling the claw, but the antennule, or “nose,” of the lobster. Hence the nose hairs, of course.


A precise zig zag pattern is achieved in Blender via dupliverts. where the hair is the object and the vert is a mesh of many verticies that specify the exact location of the center of each hair.

You are understanding the resolution/smoothness bit correctly . In Blender the smoothness/resolution of a mesh is applied to organic meshes with the subsurf modifier from which you can increase the subdivisional levels (read iterations of the Catmul-Clark subsurf algorithm) which will make a mesh denser with more and more vertices added per each iteration of the algorithm and therefore smoother . To export a really high poly count mesh all you need to do is to apply the modifier before you export (which might take a while if your mesh is extremely dense) .

So I gather from your first paragraph that you just basically need an accurate dense model to act as an obstruction for a fluid sim ? If so yes you can do this with Blender …somewhat . And if you are completely new to 3D modeling you will have as much trouble learning the toolset available in other packages as Blender .

The only thing which might seem a little problematic at first is that Blender does not have explicit unit measurements . It has Blender Units (BU) which can be defined as any type of measurement type you want, i.e. one BU is equal to one MM (sounds like a good choice in your situation) or any other measure . And you can define the scale down to 0.000 of that unit .

Actually this : might help you get started since you need to learn how to model with precision . It doesn’t deal with organic modeling per se but is a very clear guide on how to do almost CAD level work with Blender .

The only other problem you might encounter is the “while maintaining a “perfect” (read: mathematical, continuous) description of the model internally” bit . You can with some practice use subdivisional modeling to get almost accurate models but they will not be accurate mathematically speaking as NURBS modeling (you will be “eyeballing” it as it were) because the subsurf algorithm used was invented for character animation for movies not for scientific modeling . If you need an absolutely accurate surface for the simulation then this is not the method you want .

Blender does have rudimentary NURBS surfaces but its current implementation is a bit poor for any sort of robust organic modeling for most people . But in your case you might want to try them since looking at the pictures you posted I can see how you could do that using the NURBS implementation currently in Blender . You’ll just have to manually duplicate all the hairs (not too daunting looking at the pics) and snap all the points at their base to merge them …
But if you really want a robust NURBS toolset you might want to look at something like Rhino for your work .
You can do very mathematically accurate work in Blender but not with regard to precise organic surfaces … most other (animation) packages don’t either …

Good luck with your dissertation !

Thanks everyone. It seems like NURBS would be the most natural method here, since I’d like to start by modeling a smooth surface and then approximating it as a mesh, not the other way around. I also think a mathematically precise description of the surface is important since this will be used for research purposes with a computational fluid dynamics code (it is critical that the underlying surface geometry not change as I refine its mesh approximation).

If I tried Blender NURBS, could I still convert and export a NURBS model as a triangular mesh of variable smoothness?

I’m also looking into other apps such as Rhino. I have nothing against Blender, of course, I simply want the best program for the job (and it wouldn’t hurt for it to be easy to learn!)

yes. space->add->surface->NURBS Surface
Object->Convert Object Type. Click. Select Mesh. Click. Done.

Well, if access to programs and hardware is not a problem, Catia
coupled with EFD (free for school projects *) Would be the ideal combination I think.
Might be too long a learning curve for your project though.
Just a thought…

  • Maybe only Canadian schools

If you’re going to use nurbs, you’ll probably want to use another program, unfortunately, blender’s nurbs are very basic, but then again, what you’re doing doesn’t look too difficult. If you’re going to do it with blender, however, you’ll need to know how to model with nurbs, and know more about nurbs in general, than you would for other programs.

the latest nurbs documentation is here

Thanks everyone, this has been very helpful!