Good laser Scanner

[stockbub]

Any advice on a good laser scanner? What kind of quality is there? Do you use it with Artcam or what?

[Brady Watson]

How deep are your pockets?

A good laser scanner will cost you upwards of $30k and up complete - to go from physical model to 3D file with good to excellent detail. There are many cheap laser scanners out there, and they leave a LOT to be desired. I wouldn't bother with the NextEngine or others offered in the sub-$10k range, unless you are already bald. Even ones in the $10k-$20k range leave some to be desired - which is why I built my own setup (which is not for sale or up for discussion - sorry)

Just owning the equipment is one thing, but post processing the collected data requires at the minimum ArtCAM with upgrade for digitized data/toolpath reader option, and a solid handle on working with 3D within the program. Many people realize that buying a scanner, software and doing all the work is more than they want to get involved in - so they just send it out to be scanned. Unless you are scanning hundreds or thousands of items, it doesn't pay to invest in your own equipment. I know that's not what you wanted to hear...but I'll tell you how it is in reality, rather than tell you a webcam and laser pointer is going to suit your needs.

I'd be happy to give you a quote on laser scanning services if you sent me an email with photographic examples and overall sizes of your parts.

-B

[jeffreymcgrew]

Hey Brady, we've been thinking of getting a NextEngine for small scale and in-house custom work (i.e. not the large-scale stuff you do for folks). What don't you like about it? Haven't had the chance to work with one personally, so I'd love to hear your candid opinion of it.

[Brady Watson]

Jeff,
The NE is not bad for the money if doing medium sized objects, in the 4" to 8" range that don't have a lot of detail to them. The resolution just isn't there for picking up crisp details in existing reliefs. It is good for most 'feature free' models that you will scan, but the finish surface can be rather muddy. You may think off the cuff .005" resolution is good, but it's amazing how much .005" really is when you scan something with a laser. NE scans typically have an orange peel texture to them. The laser sensor array itself is of lower quality and it doesn't deal well with shiny, total white, total black or dark & light combinations on the model. They allow you to adjust laser intensity on the scan screen (by model is light or dark slider), but what if your model has a lot of contrast from one area to another? This is where the data gets ugly, as noise & data anomolies will be present regardless of the power setting on anything but a relatively homogenous surface. You simply can't paint every model to homogenize it.

While no laser will scan/pick up a perfect 90° step or edge, I found that the best the NE could do was about a 55 to 60° angle (using panel scan mode). Depending on where the model is sitting in relation to the sweep of the laser line, the laser can be blinded from the part, and geometry missed. Yes, you can stitch scans together, but you will still have spots of missing data on parts such as flat relief panels with ungulating patterns etc. I did extensive testing with the NE scanner (at an associate's place) and was disappointed with the results. The other thing is the time factor involved in going from model to GOOD scan. It can take you several hours to get a good end result depending on what you are scanning. The scanning itself is easy - you enter parameters and let it scan. The time investment is you checking the scan to make sure it is the best that you can get (and re-scan adjusting parameters if needed) and then post process the data, putting the scans together etc. The data can be rather large due to point density. You can decimate the number of triangulated points (mesh) at the expense of a faceted or reduced quality model. On a side note, Rhino will choke on many of the scans that you want to work with...it simply is not made to handle STLs over let's say 150MB very well.

Aside from the scanner, you'll need a pretty stout computer to process the scans, edit and stitch them. The software does allow you to do some smoothing and modification of the scanned result, but not to the degree that you wish it let you. The NE is a neat little device for the money. If you like tinkering around and have disposable income, then by all means get it. If you want professional results, then it isn't the right tool for the job. You would do better to buy a used Roland Picza scanner (either with 4X6X2.375" capacity or 12X8X2.375" capacity) to do smaller items with up to .002" resolution, and use the ShopBot probe for items larger than this, or just build a business relationship with someone who does scanning. I used the Roland/ShopBot probe combo for a few years before I really needed to invest in a laser to accomodate my customer's needs. A Picza is super reliable, scans overnight while you are sleeping and outputs in lots of formats. If you sniff around a bit, you'll find many used NE scanners for sale.

-B

This is a NextEngine scan on finest resolution. The background is gray (ArtCAM) and you scan see the holes in the model where the scanner didn't pick it up.



26261.jpg

Here is a close-up showing the orange peeled surface:


26262.jpg

Here is a section of the original model:


26263.jpg

-B

[jeffreymcgrew]

Brady, as always, thanks for the help.

Have you any experience with the NE's engine plugin that translates scans into Solidworks models? I have my doubts about how well that would work, but haven't had the chance to talk to anyone who's used it.

As for editing / stitching models back together, I've found that using Blender of all things has worked for very high density high poly models, but I haven't tried importing a laser scan. It works well when dealing with very dense multires subdivision surfaces and the like. I'd like to give one of your 150 mb models a try just to see what happens!

We've got a probe, older style, but it works. This is mostly at the 'see what we can do' stage, we don't have a business idea based around laser scanning. So thanks for the news about the NE, we'll look to find a used one around town.

[myxpykalix]

My limited research led me to this for more info on the NE:
http://www.youtube.com/watch?v=uXtQR7mpNZg
also I talked to the guys who have the Thinglab:
http://www.youtube.com/watch?v=854ZT...eature=related

but it was 15K!
do a search on youtube and you'll come up with lots of info.

[Brady Watson]

No problem, Jeff. I have used the SW plugin & it is a bit clunky. I cannot stress enough how you need a really stout computer to process large data sets that come from a laser - even the NE.

I've tried bringing some of my scans into Blender...and it choked. Same with Rhino on anything over 150MB, which is most things.

The SB probe is good (regardless of vintage). Like I said, you may want to look into a Picza as it fills the gap where the SB probe leaves off for smaller items. Plus, no need for powder pens etc to deal with excessive light, dark or shiny objects.

Your biggest challenge will be finding suitable software to edit your scans with the degree of control that you want. 3-axis scans are obviously much easier to deal with than full 360° ones. I've found ClayTools to be invaluable for editing and modifying 360° scans, something that is tough to do in ArtCAM or any other software for that matter - especially when the raw points file is just under 1GB.

Jack,
The ThingLab scanner looks more like a structured light scanner than a laser...as with anything, you get what you pay for, and if you really need the technology, then price is not an issue. Be careful not to fall prey to the dog an pony shows that a lot of hardware vendors will have you believe is the 'end all be all' in scanning. Regardless of the price, you will still have some post-scan editing to do with any scanner of any configuration.

-B

[jeffreymcgrew]

The smaller Picza's aren't too bad in terms of price. Thanks for the info, hadn't seen that product. I really dig the fact that it can also be a small-scale CNC, for we do have some ideas to get into tiny products, like jewelery, that just don't seem sane to try with the Shopbot (even though I feel like it would work).

I figured that the SW plugin was probably clunky. For the price I was surprised at what they were claiming it did, for I'd seen stuff like that before but only from much more expensive vendors.

As for dataset size, I know what you're talking about somewhat. Some of the Architectural models I've worked on in Revit are frighteningly large and heavy. Skyscraper models with many thousands of elements, that sort of thing. Decent 64-bit Windows computing can't come soon enough for me. ;-) Thankfully most of what we do is a lot smaller, but Revit models can really choke Blender, Rhino, or 3D Studio too, so I know the pain of having a very nice model that you can't easily work on outside of it's native format.

Thanks again

[Brady Watson]

Jeff,
I had to build a new computer (from select parts) to be able to handle the data. It's a DuoCore 2.4GHz overclocked to 3.3GHZ and 8GB of Corsair matched RAM running on XP64 Pro. I started with 4GB, but that wasn't enough! What's nice is that most hardware and software vendors are really starting to come around with full support for 64-bit. The OS started as Vista64 Ultimate, but A) I really hate what they did to the interface (like...where the heck is everything???) and B) Vista requires signed hardware drivers, which can be a nightmare when running anything but the latest mainstream hardware. Just about everything else 32-bit runs fine under XP64...Including HL2 & COD4


-B

[curtiss]

Brady,

With a DuoCore 2.4GHz overclocked to 3.3GHZ it is good you have a floatation tank to relax a little bit from time to time.

Curtis