Marthijn contacted us directly and I see in this thread you guys have been asking how the tutorial(s) are coming along. I haven't been on this forum in quite a while, our silence isn't intentional, just busy in other areas.

We have been trying to prepare a detailed tutorial with images to aid users in scanning successfully with the EinScan-S, but it hasn’t happened yet. The project we were working on that required extensive 3D scanning is no longer a priority so we haven’t been using the scanner much as of late.

Here are some quick tips though for those that find them helpful:

First, don’t rotate around your part if you are free-scanning, this will always give you a higher misalignment rate. Instead, start at a point that is very central to the part you are scanning, then pan at as close to the same viewing angle across the part at 50% overlap as one long pass. When you want to get more data from that area, start back at the beginning before you change your scanning angles, then do the same sweep again. If the data is almost identical in the overlapping regions, the alignment is almost perfect and almost never fails. When changing the angle you are “sweeping”, start at a common point and try for 100% overlap – this is very important.

Next, to reduce the amount of “stacking errors” there are a few ways you can try to globally align the scans as you move away from your starting point. Depending on the shape, you can start with a textured surface (plywood is great or something similar with non-repeating texture). If you tape or paint the base alignment area with a random pattern of dark thick lines it will greatly reduce the amount of scan data you collect to have this reference object. The random pattern also helps with quick alignment. If you offset this surface a little from your main object, EinScan will automatically remove the reference object when you complete the scan, provided none of the supporting elements are touching both are scanned, and that the reference object has less scan points than you main object. If the reference does have more points, that will be the only part that remains when the project is completed.

The flatter the reference object with relatively less depth to its texture the better it will work because it can be captured from many angles and still give back very similar data. For example, we tried the bottom of floor tiles and they worked well for that reason. When we placed our scan object on top of that, it didn’t matter how little of overlapping data we had on the part itself, or if it was going around a corner (a difficult scenario that is prone to misalignment) the scan data would snap in place exactly in the right spot because it used the reference surface scan data instead. From our testing though, wood is best.

Where this all gets very interesting is with a custom little program we wrote here. It allows you to create your own project files that will function based off a set of previous scans. So if you were to take the time to make a nice base scan of your reference object, you could save that as a project and then never have to re-scan that data. Then, once you are finished your project, simply close the program, remove the scans that were part of the initial base prep (if done properly they will have identifying info in the scan file names) and then you can complete your project, or if it is really large you can just take chunks of your aligned scan data, open just that chunk and then only add additional scan data to that region. This saves many hours waiting for alignments when scanning huge projects. We have some really big scans.

We feel these features should be handled by the software but we have met a little hesitation from the Chinese in this regard. It wouldn’t take much to get much more capability out of this scanner but we are very impressed with what we have been able to do with the investment that had to be made thus far. We have had to try many approaches though, months of effort and testing.

I hope this helps!

Best regards,

Dave @ NERV
www.NervIndustries.com