In this episode Josh chats with Michael Koper about how EINES specialize in artificial vision systems developed exclusively for quality control in the automotive industry.
Together they look at how EINES collaborate with OEMs and Tier 1 suppliers to automate inspections at every stage of the production process from stamping, painting, and metrology to final assembly ensuring consistent quality, reducing production costs, and digitizing data to improve traceability.
Quick access to recap and section transcript by clicking the arrow.
Introduction to Eines and the Interview – Host Josh Eastburn introduces the episode, featuring Mike Koper from Eines Vision Systems, and sets up the conversation on how AI-driven vision systems are transforming automotive quality control.
Mike Koper’s Background and Role at Eines – Mike shares his 15+ years in industrial automation and manufacturing, his experience with OEMs and Tier 1 suppliers, and his role as Business Development Manager for North America.
Eines’ Focus on Automotive Quality Control – Mike explains how Eines specializes in fit-and-finish, scratch, and dent inspections across automotive production, from stamping to final assembly, and why the company has become a key player in this space.
Evolution of Machine Vision in Automotive Manufacturing – Mike discusses the shift from simple part‑present checks to today’s AI‑powered, high‑resolution inspections that can measure anomalies down to a tenth of a millimeter. He recalls early challenges, like false positives from sunlight, and how modern tech solved them.
The Role of AI vs Traditional Vision – Mike highlights where AI truly adds value: accelerating ramp-up, handling product variation, and reducing false positives, and where traditional vision still works best for tasks like part‑presence and standard measurements.
Key Technology Advancements Driving Change -The conversation covers improvements in camera resolution, lighting control, AI algorithms, and processing power, all of which enable faster, more accurate inspection and broader applicability.
The ‘Digital Birth Certificate’ Concept – Mike introduces Eines’ breakthrough idea: creating a digital birth certificate for every vehicle, enabling traceability and helping OEMs resolve warranty disputes, saving millions annually. Mike explains how Eines’ systems reduced warranty claims by millions at a single plant by tracking defects back to production and preventing false chargebacks.
Future Trends: AI, Robotics, and Color Matching – Looking ahead, Mike predicts integration of AI vision with robotic repair and spectrophotometry for precise color matching, paving the way for even more automated quality control processes.
Learn more: www.eines.com
Episode Transcript:
Mike Koper
“What I’m seeing is the use of AI on maybe the most difficult applications out there. When that doesn’t work because it never worked before under the other machine vision standard, then there’s a little bit of a discouragement.”
[00:00:16.480] – Josh Eastburn (host)
Hello and welcome to the MV Pro podcast. There’s nothing I like more than hearing from people who work with technology hands-on. So, this month, we’re taking the ongoing discussion about AI into the trenches with our first system interviewer interview, courtesy of Michael Koper of Eines Vision Systems. Eines specializes in artificial vision systems developed exclusively for quality control in the automotive industry. They collaborate with OEMs and tier one suppliers to automate inspection at every stage of the production process, from stamping, painting, and metrology to final assembly, ensuring consistent quality, reducing production costs, and digitizing data to improve traceability. At Eines, the goal is to turn industrial inspection into a competitive advantage, enhancing efficiency quality and profitability in every vehicle produced. Mike is Business Development Manager for North America. With over 15 years of experience in industrial automation and manufacturing solutions, Mike works closely with OEMs and tier one suppliers to implement AI powered inspection systems. He is passionate about bridging innovation with real-world challenges on the shop floor. Please enjoy our conversation.
[00:01:24.440] – Josh
Let’s kick this off with some energy. You’ve been in this industry for a long time. What’s the best vision project you’ve ever worked on? Maybe it’s your all-time favorite or one that just really sticks out in your mind.
[00:01:35.080] – Mike Koper
I mean, that’s the most awesome question. I’m not sure I have an exact favorite. However, I really do feel humbled by the advanced vision solutions that Eines provides. For certain, when we have a customer that ask us to employ our technology on their highest volume product that they produce well over a million units a year, it’s a bit of a vote of competence that we’re able to meet their needs and the high quality. So as long as we’re delivering the proper fit and finish and build defect-free for the customer, I think our solutions then become a real integral part of their manufacturing process. And that really gives me the solace of providing something of great value for the customer, along with thinking this, when you walk away from where we’ve had our installs, on any one given day with the high volume production plants that we deal with on any given day, we’re measuring precisely 15,000 measurements, doing 6,500 specification checks and over 4 million surface images. That’s a lot of processing in a short period of time.
[00:02:55.280] – Josh
Are you able to say, are we talking about an automotive application?
[00:02:58.480] – Mike Koper
Yeah, these are on automotive vehicles.
[00:03:00.920] – Josh
Is that your primary focus?
[00:03:03.760] – Mike Koper
It is. The automotive industry has been our key target industry for our solutions. We have moved forward with becoming the de facto standard in what we call the proper fit and finish and scratch and dent inspection, along with the specification check on the vehicles, all in one tunnel, all in one solution. It’s a very complex set of many computers and AI-driven vision systems, but they’re getting the job done.
[00:03:40.800] – Josh
How did you end up in automotive machine vision? It’s a pretty niche field.
[00:03:45.900] – Mike Koper
Yeah, it is definitely. I’ve been in automotive production manufacturing for well over two decades. I led a training team through one of the OEMs in their process. See back in 15, 20 years ago, value add of machine vision, and the continuous improvement it was providing. And again, vision started off way back then as just reducing errors and simple red light, green light. It was a good part or not part. It was either part present or not part present. But now we’ve seen where it’s expanded outside of those fences into everything for vision to guide robots through interpreting different types of measure points to determining anomalies on a surface down to a 10th of a millimeter. So it has really moved. And back in those days, I would see the vision get installed, and then I would see many of the folks working in the area, or they would be calling maintenance, and then maintenance would turn off the vision. Was it quite all dialed in at that point in time? And they’re like, We still got to produce vehicles. Yeah, but the whole idea behind it is to make sure that we produce the highest quality vehicle that we can.
[00:05:18.210] – Mike Koper
So as the evolution came, in the last 15 years, I would say, has been just on a straight vertical with the applications of using machine vision for nearly every application in automotive production and many other industries, too, that have a manufacturing in their tagline.
[00:05:42.540] – Josh
I’m sure you’re happy to have the battle days in the rear view mirror, but I wonder, do you have a sense of maybe what were some of the turning points that led from that early time of this isn’t quite the stability or reliability that we’d like to offer to now where you feel like we We are locked in. This is a critical component. It’s not something that you can just switch off.
[00:06:04.300] – Mike Koper
Yeah. So what I saw the advances on is, one, the cameras themselves became much more robust and forgiving. So they had better megapixel, better number of images that they could take, photos per second, right? And that really helped accelerate the use. The software has come a long way. Back then, there were basically one or two software packages, or there were smart cameras that were all driven by that maker’s software package. Now, there are several AI-driven software packages that can control the cameras much more precisely than before. Improvements in lenses and filters, lighting, control of the lighting. That was one of the things that one of the first installs, I remember we had a challenge because every time at about three o’clock, we were getting a bunch of false positives from the system, and we were really struggling with it. And it had been due to at that time of the day when the sun would move, we would get different shadows from the skylights. And when the shadows of the skylights got on the parts, they would give us false positives. I mean, it was growth like that. Yeah, classic. It was real elementary growth.
[00:07:25.930] – Mike Koper
But now we know where to move forward and try to avoid some of those early learning errors that we’ve had. But definitely the computers and the processing has also really been a key when we could start to see that we can process thousands of bits of information off of these vision cameras in eight seconds, three seconds. That was a big move and growth in the use of machine learning.
[00:08:00.000] – Josh
We’ve been talking a lot about AI on the show this year. Really, one of the reasons I’m excited to talk to you is we’re trying to convey what’s the reality down in the trenches, not just from the perspective of marketing AI or like to be the reality, but what’s it actually like to work with this technology today? How practical is it? How really ready and reliable is it? Tell us a little bit about the perspective of Eines.
[00:08:25.780] – Mike Koper
We have a spectacular team of software engineers. I will say that. And for that, they have been able to really drive accurate and actionable program updates to include AI. And where I see that the application of AI really can be cascaded down through the manufacturing process, both upstream and downstream, is when you have the variation of product. So if the product needs just a small amount of adjustment or change in the vehicle world, you have very demanding customers. And so with that being said, you may have a vehicle that has a premium platinum package or a work package on it or a very basic package. And then there’s many trim levels in between, and all of them have different exterior components or paint on them that’s suitable for its use. Where AI comes in is as long as you can get repetition of the different types of variations, the program on itself matures, and it becomes better and better with regards to the ability to differentiate errors and defects or anomalies or mismeasurements, bad alignments, things like that are brought to the surface much quicker with AI than maybe what we used to see in just standard machine vision controls.
[00:10:06.440] – Mike Koper
So that’s nice. The ramp up-time can be shortened in terms of getting the equipment up and running and producing the data results that you can make actionable decisions on.
[00:10:23.140] – Josh
Yeah, and I imagine as a system integrator, that makes a big difference in terms of your ability to deliver faster. That all translates into your costs, your overhead.
[00:10:34.040] – Mike Koper
Absolutely, yeah. There’s a relationship to time and money, definitely, especially with new technology, right? You don’t want to be the guinea pig out there proving you on it to where it’s a little more solidified and a little more plug and go. But with that being said, it’s definitely something that when it’s used properly, it can really cut down that time to launch. And even one of the other things that I’ve noticed in the recent months in using AI with some of the synthetic information or databases with our simulation software, and that helps us dial in the equipment much more faster, too. There is two ends of the equation that AI has assisted on. It allows you to accelerate the analysis that you do in a simulation program much quicker. Even if the data is synthetic data, which is meaning if you’re giving the system what looks like an anomaly, call it a dent or scratch, it will automatically produce variations of that. It’ll produce the synthetic information where it won’t exactly look like that particular dent or ding, but it will be something just off-varied. As that synthetic data also adds to the database that you’re using in your algorithms as to determining whether you’ve got a defect or not.
[00:12:17.660] – Josh
You also mentioned not being a guinea pig for this technology. I think there are a lot of people out there who don’t really understand maybe how mature this technology is, and they’re afraid of being the guinea pig. What can you say to that? When it comes to actually working with AI tooling, how easy is it, would you say, compared to the previous generation of tools?
[00:12:41.040] – Mike Koper
It’s much more user friendly than people might think, which is when the machine vision first got going, there were a lot of programming elements that you had to be very knowledgeable of, not just machine vision, maybe even some robotic programming and things of that nature, PLC programming. You had to have a good background on a lot of different protocols and programming. But with AI, it does seem a little bit user friendly as long as you first vet out the application that you’re going to utilize it on and train that model and get one, I’m going to call it beta model built. Once the beta model is built, then the process of adding more data to it only improves that model, and then it matures to a point where the accuracy is quickly recognized and you can get there much more quicker. There’s a lot less, I’m going to call it, navigating the cameras and the set up than what they’re used to be. And I think that’s just due to the fact that you’re getting much better, I call it, analysis of the algorithms in the software, and they’re able to make offsets for the model.
[00:14:02.200] – Mike Koper
The software has become a little more intuitive, a little more user friendly than the predecessors of it, and it definitely provides the ability to ramp up your solution on the application much quicker.
[00:14:17.780] – Josh
You mentioned vetting an application. Tell me a little bit more about that, because I think it would also be helpful to maybe point out some potential lessons learned or some pitfalls that people might not be aware of as they start getting deeper into using AI. What do you mean by vetting an application?
[00:14:35.420] – Mike Koper
Right now, from what I see in the industry, the expectation of what AI can do is maybe a bit ambitious. It will get there as time evolves, but what I’m seeing is the use of AI on maybe the most difficult applications out there, and then when that doesn’t work because it never worked before under the other standard machine vision as the solution, then there’s a little bit of a discouragement. So what I mean by that is, first of all, really determine what right looks like and what wrong looks like, because that’s a real key attribute of utilizing vision solutions, AI or not. Use some real good strong simulation software to vet that application where the feasibility is determined and that the capability of the machine vision can actually do what you want it to do. I see many times where the thought, Oh, this is a machine vision and it’s got AI, so we’ll make it look at stamped metal parts and determine the scratches on the surface of it. That’s a very difficult ask of any type type of vision system. And the systems that do that are very high-end, specific towards that, and they take a long, long, long time to process the data.
[00:16:12.660] – Mike Koper
Where machine vision, the whole idea behind it is to get as quick a feedback from whether you have an anomaly or non-conformance as soon as possible. And so AI can, like I say, accelerate the launch of your solution in machine vision, but it really needs to be focused on applications that are known, that current machine vision performs, and this allows you to accelerate the implementation of a machine vision solution.
[00:16:52.360] – Josh
Another question that comes up around the same topic is, where’s the line between your traditional machine vision solutions or techniques? Things around understanding the fundamentals of the camera, of lighting systems, and then AI, and everything that you can do on the software side of things. I know Eines has a relationship with Konica Minolta, and I imagine that there’s something to be said for how those technologies fit into the solutions that you provide as well. Help me understand how you think about that. What kinds of problems should best be addressed through those traditional vision and imaging solutions versus things that you are happy to throw at AI all the time?
[00:17:34.180] – Mike Koper
Traditional vision really has worked well for standard airproofing. It definitely has found its sweet spot with regards to airproofing. Where the… I’m going to call it, I call them advanced vision solutions, where you’re doing more than just part present or color matching, or there is the right type of mirror on a vehicle for that trim level package versus a standard mirror. Maybe it’s got a tow package and it’s got oversize mirrors or something of that nature. Standard vision works great for that. It works good in the monochrome world where you’re using stereo vision to measure gaps or flushness. It works very good in that. Where the advanced vision solutions shine is when you’re trying to intricately look at surface anomalies, or pick out really my new anomalies that would determine that either a defect is there or something is not, I guess, approved through the maker’s quality standards. I can think of just if you drive your vehicle around, you look at let’s call it the mirrors. So many mirrors have painted mirrors. Some of them are chrome, some are different. But one of the projects I’ve been working on is just using AI to look at defects on those mirror caps as to if there’s any anomalies, how many different types of scratches are there in the world, right?
[00:19:22.580] – Mike Koper
How many different types of dimples of grains of sand might be in the world? This is where AI, with a very robust model, can accelerate the performance of vision. And really what it’s there for is it reduces the amount of false positives. With the AI, it can start to decipher whether or not the image has additional anomalies, and it can do it much, much quicker than ever before.
[00:19:55.460] – Josh
So let’s talk about some big wins. We’ve talked about some of the pitfalls and traps and things like that. How about the achievements that AI has unlocked for Eines, and that Eines in turn has unlocked for its customers? What can you talk about in regards to that?
[00:20:10.580] – Mike Koper
We are and have gotten great feedback on this. So with the use of our technology, which is what we call CDDS, which is like a chip denting scratch, it’s inspecting vehicles at the final point of exiting the factory floor, right? So now we have a digital birth certificate of a vehicle. But here’s a big win. Several of the plants that are for an OEM, domestic OEM, we’re getting warranty recalls from dealers when they got the vehicles that were costing north of 12, $14 million a year just in warranty recalls saying, Oh, we received your car, but it had a scratch in the bumper, or, We received your car and it had a ding, or whatever. That is all very expensive warranty work that has to be done by the dealer, but also it really impacts the bottom line of the plant, the manufacturing facility, because they get chargebacks for that. So one of the big wins is on one of the systems that we have is we’ve been able to take that data, that digital birth certificate, create a database that is able to be searched, and we can compare the VIN number of the vehicle that has been received by that dealer to the VIN number when that vehicle was on the line.
[00:21:41.530] – Mike Koper
We can look exactly where those claims might be coming from, and they shouldn’t be getting out of the factory. So we’ve been able to save just on one particular plant well over six million in cost just on that one topic of taking the digital birth certificate it of when that vehicle is made and comparing it to when it’s received at the dealer or at the customer. Because sometimes things are not caught even at the dealer, and they’re caught by the customer, and they’re like, Hey, I have two colors on my bumper, which is a recent one that just came up. Okay. We now included that as a check in our system. It wasn’t before, but in a few short, I’m going to call it keyboard strokes. Now the system is looking for comparison on coloration on bumpers.
[00:22:40.160] – Josh
I’m sorry, I missed. You said the name of that technology, CD..?
[00:22:44.480] – Mike Koper
CDDS, I call it Chip-Dent-Ding-Scratches. So we look for that on the surface of a vehicle. We’ve trademarked the name as ESDI, which is Eines Scratch and Dent Inspection. That’s what it stands for as our technology. But we’re looking at all different types of anomalies to assist our customer in saving them money in the long run.
[00:23:15.220] – Josh
Fantastic. That’s a really interesting application. The traceability aspect of what you were describing before really caught my attention. You’re able to take from recall or a customer reported defect back to the I guess the idea is to trace it back to the point in the process and then tie that into visual data that you might have from that part of the process to determine what might have gone wrong.
[00:23:39.940] – Mike Koper
Exactly. Because once the vehicle leaves the facility that it’s manufactured, there’s a lot of hands that touch the vehicle before it may get to a dealer or get to a end-user customer. If there was some unfortunate event that occurs, that devalues the vehicle itself, it needs to… Somebody’s responsible for that. That’s what the OEMs are looking to determine if it was something that they have in their manufacturing process that has this fault. And if it’s not, then how do you help the responsible party not let the issue repeat itself?
[00:24:26.840] – Josh
Awesome. Those are some great examples. What do you think is next?
[00:24:32.700] – Mike Koper
I can tell you where we have several requests as to where they’re wanting to take the use of machine vision. And two things that stand out in my mind. One is with the use of rapid computation from AI and feeding that information back to robots, we’re seeing the use of it for repairing vehicles. So that’s one thing where there’s a push in the industry to made up a vision with an actual functional repairing robot. So that real-world vision is going to be what is driving the robot, where it moves, how quick it moves, what it’s doing to make a repair, maybe put a windshield wiper on a vehicle that it was missing. There’s several applications that are out there that the use of more advanced vision will control several aspects of repair. That’s one that I see occurring. In another, I do see it, at least in my world is to utilize vision with some other technologies, like Konica Minolta produces spectrophotometers, and what those are really doing is looking at color matching. So if the color of a front bumper doesn’t match the color of the fender, and the fender was painted at the plant, but the car bumper was painted at a supplier, that shade difference needs to be identified early on.
[00:26:18.960] – Mike Koper
Where I’m seeing AI is not just the use of the AI software, but the combination of technologies of vision and spectrophotometer technology coming together and being where it can look at, call it a vehicle, it can see that there is a mismatch in color using the technology of the spectrophotometer, but doing it with not touching the vehicle because most of the systems out there right now are utilizing a touch type of application because you do really need to shroud any external light when you’re trying to get the delicateness of color matching and color harmony. So that now is going to be maybe where the use of AI can go into the software and it can modify what the spectrophotometer is seeing because it can, I’m going to call it, glean out these external lights that might be there. So the future definitely looks quite amazing in terms of where things will going with the application of machine vision. But I expect to see several types of combinations of technologies moving forward.
[00:27:41.560] – Josh
Excellent. Thank you so much for your time today. I’ve learned a lot. Where can our listeners learn more about Eines and what you’re up to?
[00:27:50.760] – Mike Koper
Yeah, if they just do a quick search and go to www.eines.com, go to our website. On the website, you’ll see many of the challenging applications that we’ve been involved in the past with our customers, along with knowing some of our what we call dedicated solutions that are specific to what our customers need. There’s just small amounts of customization, and those systems get launched relatively quickly. We’re always looking for and open to challenging our team, our application software engineers, our technology. We like to align with customers and put it in our innovation lab and try to get them the best results to fit their needs.
[00:28:41.160] – Josh
Ready to get your hands dirty, huh? That’s great.
[00:28:43.560] – Mike Koper
Absolutely.
[00:28:47.320] – Josh Eastburn (host)
Thank you, Mike and the EINES North America team for putting this together. As mentioned, you can see more about EINES Tailored vision solutions for the automotive industry at eines.com
[00:28:59.560] – Josh Eastburn (host)
Now, Well, let’s talk about your specialty. I’d like to fill up our calendar with more real-world stories from tech pros. If you are a system integrator or vision systems engineer, please reach out with your unique story. We want to help people understand not just the hype, but the reality of working with modern machine vision technologies. You have that story. So please share. You can reach me at josh.eastburn@Mvpromedia.com, or find me at me on LinkedIn. For MV Pro Media, I’m Josh Eastburn.
