Smart quality control: How machine vision optimizes production

Simon and Matthias with a small part of the tape.

At our customer’s facility, where kilometers of plastic adhesive tape are produced daily, quality isn’t a detail, it’s essential. A single small defect can render an entire, expensive roll unusable. When the customer started looking for a way to detect defects early and accurately, they turned to Heliovision. The result? A powerful visual inspection solution that not only detects defects at a speed of 1.5 meters per second, but also precisely pinpoints their location on the 4-kilometer-long tape. We spoke with Matthias and Simon, two of our engineers who led the project from start to finish.

From Proof of Concept to production-ready solution

“The first step was researching the right camera technology and lighting setup,” says Matthias. “We had to determine whether 2D would suffice or if we needed to switch to 3D cameras. In the end, a smart 2D setup with two types of lighting proved to be both sufficient and efficient.”

Simon then took on the algorithmic part: “My focus was on writing the code to detect defects, such as missing plastic or irregularities in the wires. We used alternating lighting every 400 microseconds, the lighting switches, allowing us to detect different defects using just one camera. Pretty impressive that it worked so well.”

Challenges and creativity

As with many R&D projects, it wasn’t always smooth sailing. “We tested several light controllers — one of them literally exploded,” Matthias laughs. “But in the end, we even developed a new feature together with the supplier, specifically for this project.”

Defect positioning on the roll brought its own challenges. “Because the product is so long, we had to pinpoint exactly where each defect was located,” Simon explains. “We combined data from encoders, buffer logic, and machine signals to map that out accurately.”

Why not AI?

In a time where AI is booming, we consciously chose a white-box approach. “The client wanted to be able to set parameters themselves,” says Simon. “That’s not so easy with an AI model. Also, there wasn’t enough faulty data to properly train an AI. Traditional algorithms allowed us to work faster and more transparently.”

Impact on production

The new visual inspection system offers several direct benefits:

• Less waste: Defective rolls are detected earlier, before they reach the end customer.

• More efficient workflow: Operators get a clear overview of where issues are located — no manual inspection needed.

• Improved quality: Continuous monitoring helps raise overall product quality.

Simon adds: “We even detected an error the customer initially didn’t consider a defect. After a closer look, they realized it actually could be a problem. So our solution is more precise than what they were used to.”

First project, lessons learned

This was Simon’s first project with us. “Very exciting. You learn what it really takes to get a system running in production, how to communicate with the client, what’s feasible and what’s not.”Matthias adds: “We learned a lot about lighting techniques, synchronization, and collaboration. Having two people on a project not only brings more ideas — it also makes it more fun.”

In Short?

A successful and challenging solution that does exactly what the customer envisioned.For Simon, it was a valuable learning experience that once again highlighted the importance of lighting in computer vision setups.

Curious how Heliovision builds smart vision solutions for industrial environments? Feel free to get in touch.