If your company is considering or is already Reverse Engineering models from scan data (point or mesh) or using scanners for verification, we would love to hear and learn from you. Just drop me an email at email@example.com with an introduction of your company, RE applications, workflow etc.
Manager, Solid Edge Planning & Strategy
This is sort of a "chicken and egg" question. I would love to use "scan data" or a scanner, but without an easy (read cheap) way to get that into Solid Edge, it's perpetually on my wish list.
Laser scanning is a can of worms. To properly process point clouds for an undetermined result is not worth the software costs, required CPU cycles, and learning curves.
When starting a laser scanning process of any kind, the required results must be VERY specific. One step out the what was originally specified.........another $30K or $60K software and 3 months learning curve is required.
The one thing most people don't get about it is this: Consider scanning a single dice. This is typically done by scanning from the view of 8 corners. Most Geometry requires 16 scans. All 16 scans must be loaded up at once for alignment and segmenting. Most scans are 1,000,000 points at a resolution of 1/10,000 of the scanner range.
You better have a computer that can load 16 million points or your work flow is screwed.
Once you get past the registration and segmenting process (that is result specific) then the kind of required geometry determines the software:
Polyworks: Surface polygons * non registration cloud alignment
Geomagic: Enclosed Polygons
Cyclone: Pipe and steel shapes (And they stole Polyworks could alignment, but does not work as well)
Downstream processes are very specific to required results.
To do "anything" you need about 3 years learning curve and $150K software.
Oh and the fastest possible machine you can get. (Period)
We had a couple of boats scanned this Fall. On one of the boats, we wre checking the keel and rudder alignment with the hull. On the other boat, we were scanning it to get the as built shape so we build some tooling so that we modify the transom on the boat. We hired a guy with a FARO scaner and he was there for about 4.5 hours (3.5 hours on the whole boat scan for the alignment check, and another hour to scan the transom area). I think the FARO scanner was set at kind of a low/medium resolution and was recording something like 300,000 points per second. He processed all the information and only charged us $1,500. He gave us RXXXX files but could have given us STEP or IGES files just as easily.
A scanning service is one thing. Doing it your self is another. Letting others deal with the headaches is the right way to go if you only need a few things scanned.
I'm gonna switch hats for a second. This is one of my pet topics as a user. In my previous position I was a product development consultant, and did a fair bit of scanning and surfacing over scanned data. Since I did my own scanning, I knew how painful a lot of this work is, and how much it could improve. I was using software included with my scanner (Next Engine - laugh if you will, but you should see some of the stuff I was able to get done with that thing), and the point cloud software included with the top level of 'Works. I did everything from civil war gun reproductions, racing seats, baby dolls, packaged food, to body parts, scale models, shampoo bottles, equestrian helmets. The tools I used were weak, but I was able to get decent results just by spending too much time on tweaking the final shape.
I think the first thing a CAD system needs to be able to do is to lay sketches over point cloud data by two methods: "live section" type where you place a plane and the software gets the closest spline to the cloud points on that plane, and second by just picking curve (spline) points manually on a plane and referencing the scan points. You also need good visualization of the point cloud to do that, so everything you can do with a section of a solid, you need to also do this with a point cloud. Visualization as a subdivision surface would also be cool.
It would be very cool to have something like TSplines that converts mesh to nurbs patches, but to do that, you also have to have good smoothing tools, and even point editing. You can't just assume the cloud is correct. I've never seen a situation where the scan data was 100% reliable.
A reasonable set of comparison tools would also be important because this stuff is needed for QA/inspection as well as product development.
It would be cool to be able to interface the CAD directly to an arm/probe or various light/laser scanners, but that might be best for a partner product.
Mesh data is going to start coming from more and different sources as 3D print becomes more prevalent, and people are looking for data. 3D scan is also going to increase, and scan/cloud/mesh/subdivision/tesselation is going to become another data type just like IGES, STEP, and Parasolid that CAD systems have to be able to contend with.
There is also the fact that we could use data from all of those subdivision modelers as reference to model around or for renderings. Plus, the world of 3D medical data from various types of scanners is incredibly important in medical device development.
It would be ideal to have editing capabilities like Mudbox/Zbrush/modo/3dsmax/etc within the CAD tool.
It sounds like tracing is your desired method of reverse engineering. Couldworks is an add on to AutoCAD for Cyrax data for that exact solution. I have been out of that game for a long time, Cloudworks may already exist for some of the parametric modeling software. My first guess would be inventor.
Inspection and interference checking is one of the best uses of scan data because the point clouds can be used in raw form. Also you can decide how much time you want to spend cleaning and segmenting point clouds.
Turning points into 3D models is where things gets difficult. Tracing is the only place one can force point clouds into vertical and horizontal planes without manually creating new planes to replaces planes calculated from clouds.
@MLombard My guess this is a teaser or at least a beginnings of a market analysis of maybe bringing some of the Siemens PLM Imageware capabilities into SE! That would be absolutely fabulous.
Link to Siemens PLM Brochure- which needs a serious facelift since it was last done in 2007 and still have CRT monitors and aging fashions! ;-) If you want to present this as a current and updated software...well, then sell it like it is one!
Right, we want to get some deeper insights into how scanned point & facet data would fit into a CAD workflow. As you can see here, that is hard to pin down with a plethora of data sources, software tools and expectations. Thanks for your emails and discussions here that are feeding into our 'figuring it out'.
RyanM wrote: ...which needs a serious facelift since it was last done in 2007 and still have CRT monitors and aging fashions! ;-) ...
Yeah, as soon as they put me in charge of fashions... Corporate sandals and cargo shorts with earth-tone henleys!
Thankfully, we've probably got a better shot at point cloud editing than corporate sandals.