Reverse Engineering and Point Cloud Data
When I was doing a lot of contracting work, one of the things I wound up doing, somewhat reluctantly, was reverse engineering. I had an inexpensive laser scanner with horrible software, but I made it work. I had projects that required scanning every thing from human feet to race car seats, baby dolls to civil war rifle parts. Quite often I had to fit history-based surface features over the resulting point cloud data so that I could get a fully editable model as a starting point. The only project I ever went back to the customer and told them I just couldn’t do it was scanning that baby doll. The scan just didn’t have the resolution, and was highly sensitive to small face incident angles. I know more sophisticated hardware and software have existed for this sort of thing for years, but for a steep price, and if you were going to get into that kind of thing, you had to do a fair amount of it to justify the cost.
3D printing is reaching down to the retail level, and what kind of data does 3D printing use? That’s right. Point cloud data. So we have this burgeoning market for this type of data, although it’s probably a give-away sort of market, it will at least create some demand for tools to manipulate what is often thought of as static STL data.
Combine again medical geometrical data, which is also stored as point cloud. Data from body parts, bones, internal organs, teeth. All of this stuff finds itself being used in engineering applications, where CAD tools have to interface with that data. Even FEA and video display make use of some form of this method.
Last but not least, think of all the shared 3DS (3DStudioMax) models out there from graphic artists using a huge range of mesh modelers and editors. Little of it probably has any real engineering value, but think of it as 3D clipart, maybe for a size reference or for a rendering prop. This bonanza of geometrical data puts CAD vendors in a great position, potentially speaking, to make use of it all in the real world.
There is a great gulf between point cloud data and NURBS data, which some 3rd party vendors like modo and Tsplines have started to bridge. And then there are specialist high end applications that have been around for some time, like the Geomagic tools.
I expect sub-division surfaces techniques will enter the fray, but I don’t expect they will take over totally.
In Dr. Versprille’s guest article on the Solid Edge Community, speaking of the future of NURBS, he wrote that “ I expect sub-division surfaces techniques will enter the fray, but I don’t expect they will take over totally”. I share that view, and would like to see more point cloud tools showing up in the mainstream CAD tools. Engineered organic shapes are going to increase in the future. For prismatic parts, point cloud isn’t all that practical, but for organic shapes, mesh is really hard to beat. This is valid data which has applications everywhere you can use other CAD data, not just for creating game characters anymore. We need real tools to work with it.
Maybe I’m not reading the comments from everyone correctly, or misunderstanding something, but you all seem to have missed the great Sub-D surfacing integration in the recently announced NX9. They seem to have one-upped just about everyone else with they way they integrated it (Including manipulating the sculpting with the new Leap Motion controller). Looks very impressive. Just go search “NX 9 Sub-D surfacing” and watch some of the You Tube videos.
I was talking to a sheet metal guy recently about changes to SE sheet metal and he had some interesting comments that apply here. One of the simple things they (sheet metal guys) had been after for some time was finally done. Later Bill McClure mentioned to him that this particular project had cost about $100,000.00 to do. It shed a light on some things for me and explained why even though I would REALLY like to see an interface for my Faroarm into SE I probably never will. The percentage of users that actually do this kind of stuff is probably quite small and there are probably hundreds of things that will be done before this is considered if it ever is. Now it would be glamorous to put this in there for PR’s sake, assuming that SE has a PR department that would know what to do with the information, but practically speaking would benefit very few I am afraid. SE has made no bones about becoming the best for the MCAD market in general first so I don’t look for these point cloud things any time soon. It is why outrageously expensive programs still survive for reverse engineering.
Sometimes I think people forget that SE works on larger objects, too. We need a data gathering solution for objects the size of boats, trucks and buildings that doesn’t cost an arm and a leg. It doesn’t need micron precision, but we do need accuracy to check models against real existing objects. For example we are doing a 5 floor railing for a central building stair installed on another company’s steps and stringers. Field measurements on this are going to be tedious and error prone. 1mm precision would be more than enough.Does anyone have a suggestion?
Larry,
You can get hardware that mounts on a speaker stand or that is handheld. A company up in the Maryland area has scanned a lot of the national monuments in DC. I don’t know about stuff that big that is affordable or what range you’re looking at, but you might consider eBay. I’ve seen some good condition scanning equipment that was very reasonably priced. Remember, you don’t have to scan the whole thing. You can just scan a repeatable unit and pattern in CAD.
Roger,
If I knew anything about the future direction, I wouldn’t be able to talk about it, so ignorance in this case allows me to offer an opinion, or a wish in any case.
Matt
I guess I found this comment in the paper a little strange (and disappointing) as the integration of sub-d would seem of obvious benefit. Maybe they have something even more cool in mind!
A couple of weeks ago a client asked if I could create a part, a guitar bridge with organic forms that would be CNC’d. I played about with it a bit but doing it in SE was going to be a challenge and the cost proved prohibitive. Now, this is not my normal sort of work so I didn’t loose sleep when I didn’t get the job, however it did get me thinking.
Today I thought I’d give Fusion 360 a go now that they have integrated T-splines (and it is still free!) to see if it is more user-friendly than Modo. First off the web site is superb and I was up and running in a couple of minutes. Another couple of minutes (without consulting the help) I’d imported the clients artwork and was tracing the outline with the spline tool. The spline tool can look a little “busy” at times but I achieved the shape easier than in SE. I only had a short play with sub-d but it was easy to get started and the interface was excellent.
I don’t get the impression Fusion would be much good for general mechanical design (which is primarily what I do) but it could become the “go to” for complex shapes.
The point I’m making is that Autodesk may just have made complex modelling accessible to the masses with this product. The interface is excellent, direct distribution has you up and running in minutes and it is priced low enough for the casual user. I just wonder at the wisdom of Seimens disinterest in this area.
Roger
Roger,
Yeah, I hear you. I agree. It could be that synchronous surfacing will do some of what subd modeling would do, but it’s just conjecture at this point to talk about either. NX has synch surfacing, but I don’t know if it is at all geared toward working with general point cloud or reverse engineering. I think NX has separate data acquisition/reverse eng tools.
According to the properties of that PDF, it was written in 2009. I think Tsplines was around then, but not Fusion360. The big 3d print push was still young then too.
It’s hard to say what will happen in 5 years, but I’d like to see “general purpose” CAD that could handle machine design and product design.
Matt
If the paper was 2009 then the views expressed could have been formulated in 2008, at which time both laser scanning and 3D printing were less accessible. A lot has changed in 5 years and maybe Siemens view the market differently now – who knows?
Sync surfacing sounds interesting and the NX video of the camera (I’m assuming this sync surfacing) is cool but I wonder can it give the sort of control of sub-d/T-splines. I’m not much of a surfacing user so am pretty ignorant but today when using Fusion, the ability to make an edge along UV line was impressive.
Anyway, it’s an interesting conversation and when you see what Spaceclaim have done with point cloud in 2014, I reckon it could quickly become a differentiator.
Roger
Larry,
There are quite a few contracting companies that do what you are looking for. The type of scanning for large subjects is LIDAR. These companies will scan entire buildings, caves, mining operations. The accuracy of the scan depends on the environment being scanned and whether or not they use calibrated reference spheres to keep everything to a known scale.
This is a paper I came across last year posted on Evan Yares’ blog http://siag.project.ifi.uio.no/problems/allen.pdf, written by Siemens Chief Technologist, George Allen. It is an interesting read.
The summing up on page 9 may give an insight into where Siemens are going with their research and that doesn’t seem to include sub-d.
There is no date on the paper (though it looks fairly current) but I guess this sort of thing is always up for review as trends change.
Roger