SPECIALIZED ON SOFTWARE FOR GEOMETRICAL SHAPE ANALYSIS AND MODELING OF FREEFROM SURFACES.
Cyberstrak offers a Plugin for the CAD system Rhinoceros 3D. It contains functionality to support designers and stylists on their daily job. Focus is easy and intuitive usability as well as compatibility to the existing CAX products on the market" hmmm
@Peter_Salzman If you can add mesh editing and 3d scan point cloud or mesh alignment to this, and enhance these reverse engineering type tools over time, then imma need you ta take my money and gimme a license
It’s surface modeling software aimed at classic patch modeling, not RE software. I’ve used VSR for years doing RE work on complex laser scanned data (for example, an entire Boeing 777) and some of that same capability seems destined for Cyberstrak, but neither the purpose of VSR or Cyberstrak was to ever serve the same purpose as RapidWorks (er PolyWorks?) or NextEngine.
mostly just need mesh-alignment algos, cause that’s basically the only thing missing in Rhino relative to RE imo.
yeah, everything can improve over time, but there’s still a major hole in terms of whether Rhino can align meshes or not… including point-clouds of course.
That type of capability was never a part of VSR, and I don’t think I’m speaking out of turn in saying that Cyberstrak will not have that either. That’s really laser scanning post processing software, not modeling software. In VSR, the implementation was that you could use a mesh (or a surface, or polysurface) as a reference - Projection Base in VSR speak. This allowed you to “paint” NURBS patches on top of a mesh. You could even control point edit those surfaces, and “re-snap” them to the reference. You still had to have a good understanding of patch layout, control point modeling and matching to make a model with it, but you could get from clean/aligned/unified laser scanned data to a very high quality NURBS model if you knew the way. That type of workflow and doing assembly/alignment of individual scans are on totally different planets.
Dwelling of whatever VSR was in the past, will only lead to obsolescence.
Laser scanning post processing etc, is modeling software. To think that it shouldn’t be, is to also lead down a path of obsolescence.
I would agree however that their is a trend of post processing software that lack many “modeling” abilities, due to the common ideology that there doesn’t need to be any.
Hence, imo the niche yet to be filled of said ‘mesh/poincloud alignment’ algos.
That’s literally RE. Hence, it’s very existence highlights the lack of alignment algos.
That’s a round about way of saying, “modeling” and “RE” should be hard to do or nearly impossible for newbs to do. Which is silly cause that way of thinking is pretending that obsolescence doesn’t exist.
I suppose we should go back to using pencil and paper to make drawlings instead of using CAD’s at all.
They’re not on different planets. lol. They’re literally on the same planet
I know I’m pressing a stressful point. So don’t get me wrong, I’m just coming from a stressful background.
And I know I’m kinda obligating Rhino to have a solution, but that’s just my angle. I don’t mean to frustrate anyone too much
I know I’m suppose to use meshlab or something but I don’t wanna, I’d rather use Rhino
Imagine if Rhino could align meshes and pointclouds, and maybe even nurbs and subd’s
If someone said to you “I wish my car could also be a dishwasher,” you might start to scratch your head and look at them a little funny, no? Just because modeling in the case of RE is downstream of mesh alignment and registration, doesn’t put them in the same conceptual box. So, okay, maybe on the same planet, but definitely different things.
MeshLab is open source - you’ll have a far better chance of success on this front if you engage with that community and see if there is any way to implement MeshLab tools into Rhino as a plugin. That would be totally excellent in fact.
Concerning Scan alignment, what exactly is needed? Do you talk about scans of the same object coming from different scan processings that need to be adjusted to each other? With overlappings and different 3d orientation and stuff like that? Just to understand the problem…
Well, I’ve been a part of the 3D scanning industry for a fair amount of time now, ever since about 2006-2007ish.
During this time I’ve learned alot about what it takes to obtain high quality digital models of objects which are to be scanned.
Unfortunately there’s been some odd shifts in the industry, and I would say that alot of the options that users have these days, aren’t going in the right direction – imo.
However, relative to the direction I think it should go, I believe that the scanning process should include the ability for software to have an algorithmic stage in the workflow whereby the user can spend the necessary time to align scanned data prior to the following stages where the scans become merge/fused together.
Otherwise what culminates in the workflow is compounding errors.
And this is the misfortune of the fallacy that the industry is leading everyone. The erroneous belief that 3d scanners should give the user no other options for alignment but to rely on the bad sofware algos to align aimlessly.
Users in the 3d scanning industry are being ever more lead down a path that it’s ok to be moving the scanner around all the time and relying on the software to automatically align everything behind the scenes.
The result of todays trends, are simply very bad quality scans.
Yes, and no. The algorithms I believe are possible, don’t necessarily need to have any few meshes make sense to eachother, but to merely fit them the best as possible.
In otherwords, when scans do correlate and are of the same subject, then yes they’ll likely fit nicely, while also users could theoretically use the same algos to play around and fit completely unrelated objects together in a best fitment scenario – per say.
Yes. Of course overlappings etc, will most likely include erroneous data that should otherwise be trimmed prior to alignment algos and be reduced in error as to prevent tolerance deviations etc. in earlier stages of the workflow.
I’ll try to add a few links, where I’ve tried to distill down these theories as I see them:
With the right equipment and software, this is how high end scanning outfits currently operate. I’ve worked with Mimic3D over the past 12 years or so doing full airframe laser scans. They are absolutely cleaning and aligning scans on the fly, over the course of the scanning window. We can tell for instance, when a scissor lift’s cylinder seal has failed, because the sag over the course of the scan will cause the alignment to fail. There is a ton of room on the back end to massage alignment if needed. On large objects we are using a combination of spherical tripod scanners in conjunction with laser trackers and high quality reference targets. On smaller projects, the laser tracking aspect can be omitted so long as there is sufficient reference(s) to register to. The end result of this is scans that are of the highest quality and accuracy, and are simply a joy to work with. Nearly everything you describe is in my opinion the downfalls of cheap, handheld, prosumer grade laser scanners. In laser scanning, you get what you pay for, and for me, the recommended vendor list for any large scanning project where quality and accuracy is important is a list of one - Mimic3D. As you say - stitching together 4 dimensional space time is no insignificant task (for instance, there is a true art to scanning something as long as a 777 wing, which can move significantly over the course of a scanning period by temperature swings alone), but I can tell you that it’s certainly possible, with the right tools and talent.
This is the result of hundreds of scans being stitched together in post (but checked/cleaned and culled on the fly), but in the final mesh result, it’s impossible to see the overlaps in the scans - one of the true signs of a properly done scan job:
It’s seamless, accurate and has an uncertainty that can be calculated both locally and over the entire length of the airframe. For instance - we found the fuselage was about 0.25" shorter than the Boeing spec, but the temperature in the hangar was a few degrees under what Boeing keeps their assembly line at. Back of the napkin math showed that using the CoE for aluminum and the temperature delta would have put the model spot on to the Boeing specced overall length.
They’re not perfect, and the errors can propagate. Maybe your equipment is different and probably cost over $20k-$70k plus annual fees etc.
Hmmm, yeah scanning objects that move is no good. I’d almost want to scan the parts that don’t move, and then superimpose them in the CAD later in the workflow.
Yeah that’s pricey.
That sounds closer to the range I’m used to. At the smaller scale is probably where you’ll see those “on the fly” algos propagate more errors especially under certain movement and lighting conditions.
But there’s pros n’ cons, even the large curvature smooth airframes seem impossible without super duper extra equipment or thousands of tracker sticker dots
Yeah that sounds huge. But I don’t see how thermodynamics would effect that more than the error propagation I’ve mentioned. I would think that the thromodynamics would be more stable. But I guess it depends on the time of year, day, location, weather etc.
I would also thiink the software and hardware has a harder time aligning huge smooth surfaces without all the extra special equipment. Sounds really pricey, probably $100k’s
You’re probably short a few hundred scans, depending on the surface area of each scan
I see some anomalies in a few rivets in there Beautiful mesh though
Well my data archives I’m mostly concerned about atm, are about tennis shoe size objects, and maybe a few automobile size objects. But no 777’s or 747’s yet
For the shoe size objects I will want an alignment of like +/-0.005"ish, and automobile size one’s maybe 0.02"ish.
I definitely plan on trying out meshlab for that data. I’m just not sure if I’ll have file format data lock or something.
It’s possible Rhino will be able to convert it, from like .scn to .stl or something. It’s been a while since I’ve checked the format ext’s
Mostly I’ve been ranting over the yrs about the technologies that are $30k and cheaper, ever since my data became obsolete and I can’t access the original softwares.
I feel like anew brand needs to take the torch, but who knows when that will happen if ever.
Maybe someday I’ll try a $75k creaform version, but I’m just not impressed with their $30k version.
The cheap white-light tech, prob just not as good as those expensive blue-laser versions
If you want to scan an automobile (which I’ve done plenty of as well), then proper gear (currently) is an Ultra Short Range Surphaser. You don’t need a laser tracker, you just need the car on jacks and some reference objects on the floor surrounding the car. You need to make sure that 3-5 of these reference objects are visible in each scan for registration. You can hire this work out - the cost should be in the low thousands of dollars, plus travel if applicable. Find a vendor with a Surphaser, ideally an ultra short range one but the mid range models will work in this case too. If you’re doing smaller objects, your money is far better spent mailing the objects to Mimic. I don’t have any financial stake in Mimic, they’re just the only people I know who do these jobs right. My experience has shown that the vast majority of people are better served hiring out laser scanning than owning the equipment. Unless you are scanning as a business, and business is good, you’ll get a far better product, for a fraction of the cost by using a good vendor. A good vendor will have a wide range of scanners, and can pick the most appropriate unit for the job. Don’t spend $75k on a Creaform unit unless you actually are running a scanning business. I’m saying this as someone who once spent $55k on a Creaform unit. You can also send files to Mimic for post processing - I’ve had instances where clients decided to use a different vendor, and the first thing we do is send Mimic the data to get them to try to salvage the junk I’ve been sent.
And to add - nothing that Cyberstrak is or could possibly do will solve the inherent issues with cheap (by cheap I’m saying under $100k, but ESPECIALLY under $50k) handheld scanners.
hmmm
Hence, it’s very existence highlights the lack of alignment algos.
I could refute that but that too funny. I’ll give you that one. 
