null
Introduction to 3D Scanning Technologies and How to Choose Them

Introduction to 3D Scanning Technologies and How to Choose Them

3D scanning is a hot topic, and there are many different technologies to choose from. In this podcast, Len Chamberlain, CCO, and Dan Howe, Regional Development Manager, from LMI Technologies introduce to you the main types of 3D scanning on the market, and which ones are best suited to your needs. Listen now on YouTube or read the transcript below.

Podcast Transcript:

Hello, I'm Darryl Seeland, Editorial Director of Quality. I'm here with Dan How and Len Chamberlain to discuss 3D scanning. Dan is a Regional Development Manager for LMI Technologies, and an expert in 3D machine vision for factory automation applications, including a focus on laser line profiling, and structured light techniques for automation, quality control and process optimization. Lennar into mechanical engineering degree from the University of Colorado, and spent the beginning of his career designing medical equipment and production processes for Gambro. And Medtronic, when Geomagic was purchased by 3d systems, he became the VP of sales for the entire software division of that company. Thanks to you both for joining me today.

Can you both tell us a little bit about your experience and your background in 3d scanning?

Sure. I started with LMI Technologies back in 1998. I've been in this industry for about 23 years. And some of the first stuff that I did was actually related to the transportation industry looking at 3D dimensions of pavement, which is actually something that's used in assessing roads, looking at texture, rough shape, things like that. But over the years, that's expanded into all sorts of industries. So it's, it's really fun to see the different applications the that 3D is applied to.

I started about the same time, I think 1996, working with laser based triangulation scanners, primarily for reverse engineer and offline inspection. And I've worked both in the hardware industry in the software industry. I started with LMI about seven years ago, and was really attracted to taking the 3D scanning technology that I have used offline and looking at inline solutions, something that could really be deployed and mass and and be effective in terms of helping production lines gain greater efficiency and greater quality all as an output product.

What types of 3d scanning methods are there out there?

Well, there's a lot of different methods that people may be familiar with, some of which are kind of more long range. And some are shorter range with higher resolutions. Only a few of those are typically used in your quality control manufacturing environment. But it ranges from things like LIDAR, which is a large range scanning technique using time of flight with pulsed light, which people might be familiar with that might be used in your self driving cars, as well as terrain mapping, things like that. There is a technique called pseudo random dot, which is typically a dot pattern that's projected using infrared light, that's a technique that is a low cost low resolution. So typically around five millimeters to 10 millimeters that's used in a product like the Microsoft Kinect, which people might be familiar with, with the gaming platforms. But then when you get into inspection, you're typically with things that we use, like fringe projection, which is a structured like technique for area scanning, typically on the order of a meter or less in size of your object gets you down to a sub millimeter. And we use a lot of line triangulation. So that's laser line with cameras, and you can get down into the 10s of microns, or even single micron resolutions, all the way down to techniques like con focal, which can be sub micron. And even interferometry, which is something that we don't use, it's typically something you use in the lab, that'll get you down into nanometers, even angstroms.

As Dan mentioned, it's a pretty broad range when people talk about 3D vision and 3D scanning. LMI's focus and the focus of this talk is really the technologies that are most applicable to the production line things that are capable of high speed and high accuracy, and which industries would use these 3D scanning and measurement methods.

There's a big range. We started a long time ago in the lumber industry, believe it or not, so people that will scan a log to determine its shape and figure out how to cut that log to get the most yield, as well as scanning boards. But it ranges from things like automotive looking at sheet metal to engine components into rubber and tire there's actually quite a bit of measurement done on the tire building process, all the way through things like food, packaging, but then you get down to real high resolution stuff with consumer electronics. So things that are on cell phones or tablets, or measuring products like medical devices. So it can be pretty much any industry you can think of, we really ride the same wave as automation does in general. So where where you find automated production lines and automated processes, you find 3d machine vision and 3d inspection in line. So it's really an exciting place to be. And it's one of the reasons I really enjoy this job is that there is such a huge variety of applications that we address of all of these 3d scanning methods you just mentioned, which works best for industrial factory applications and why.

So typically, in factory inspection, the techniques that are used most is fringe projection, laser line triangulation, and confocal. Technology. And that's basically because they have the types of resolutions that are required in the factory inspection applications, as well as the speed. Fringe projection is typically used with stationary parts where there is no relative movement between the sensor and the object being inspected, that can also be used in an end of arm robotic application. Line triangulation and confocal are typically on a conveyor, so you need some relative movement to pass the part underneath the sensor. And you're taking those measurements as the part passes.

And how does one choose the best technology for a specific application? 

A lot of it starts really with testing, using different types of techniques, different sensor models to determine which one actually does the job. But one of the other criteria is understanding the cost of the different techniques and making sure that you're matching up the required resolutions with the budget to get the proper return on investment.

Most suppliers in this space will allow the production line to evaluate different sensors. I think that's an important piece because it's hard to have an innate feel of what sort of resolution is really required to detect the defects that you are finding in your parts. I would always recommend working with working with somebody who's well familiar with the space and getting an eval for your for your particular application. Besides that, besides the cost, I think speed is another one that we hit upon earlier, your production line speed is going to drive significantly, what you're able to measure and what sort of computing power is needed to be able to process the data at that rate.

How do users use these these sensors? How do they handle the massive amounts of data that's going to be collected?

3D sensors are only as good as what you can do with that data. It's critical that they have software and measurement tools that go with it. And, you know, we provide what we call smart sensors. And these have onboard processors that do edge computing. So you might get from a scanner, millions of points per second, produced millions of 3D points. And you don't want to necessarily pass all that data off to some computer sitting somewhere, because it's very difficult to do. So when you're measuring something, you make gather all those points, but you're really only interested in a few critical dimensions of that component. And with edge computing, you're actually doing this processing on the sensor itself. And so determining from that data, what measurements are important to the customer, maybe you're finding the position of a hole and understanding its position relative to some other feature on the object. And the sensor itself can provide either that measurement value, or perhaps even just a pass fail, where it's just sending 'this is good' or 'this is bad'.

Now many years ago, when I first ran across this technology, I thought, well, gosh, it's going to revolutionize the industry. And some 20 years later it is it is now starting to happen. I think it really had to do with the the ability to process data where the data was taken. So the edge computing the Dan mentioned, we've got applications where when we're scanning large parts at high resolutions, it's easy to generate, scan files that are half a gigabyte per part. And that's not something that you really can manage if you've got to pump that through the factories network in order to get a decision back. With the advent of ARM processors, inside the sensors themselves, it allows them to make those pass fail decisions right on the line only save the data that is really relevant to that process and keep the production line moving at the at the pace that's expected. So it's it's been a real transformative event for us

Is artificial intelligence and deep learning being used in these applications?

Yes, it's beginning to that technology has hit a point where it is very viable in a lot of different industries and certainly in the machine vision industry. Depending on the type of feature you're looking for, it may be required. We do have a lot of measurement tools that we call, you know, rules based so that's really setting up tools without artificial intelligence and that's typically looking for features positions of features. But when you get to defect detection, things that maybe aren't necessarily is easily defined as to what makes it good and bad. That's really where artificial intelligence and deep learning come to play where you can teach, teach the sensor what is good, what is bad in AI figures out what criteria defines that

Correct, there are applications that are outside strictly speaking quality there as well. So where you're classifying certain types of objects might be grading meat or grading lumber for example. But on on the pure inspection, front of artificial intelligence tends to be replacing more human operators that are looking for cosmetic defects something that is harder to define on a CAD file for example, but it is a it is a tremendous area of growth and and really broadens the applications for which 3D sensors can be used. I'm particularly excited about the intersection between artificial intelligence and the use of 3D data. That is that is a space which is really early in its development, but has a lot of potential.

Thank you both very much for your time. We really appreciate it. 

Apr 1st 2021 LMI Technologies

Recent Posts