Danube Dynamics, an Austrian embedded-systems developer, has built an AI-powered inspection system for access-technology manufacturer EVVA Sicherheitstechnik that cuts manual quality checks on highly reflective locking components from more than 30 seconds per batch to under five seconds. The system combines a closed test box, four independently controlled RGBW lighting bars and an IDS uEye XCP USB3 camera, model U3-36P0XCP-M-GL Rev. 1.2, to catch surface defects smaller than 0.1 mm on metallic parts where glare and reflections routinely hide flaws from cameras and human inspectors alike.
The camera is built around a 19.80-megapixel AR2020 sensor from onsemi, delivering 5K UHD resolution at 5136 x 3856 pixels on a 1/1.8-inch rolling-shutter sensor in a 4:3 aspect ratio, paired with an 8.5mm IDS lens, type 20M11-C08528. The system photographs an entire 400 x 300mm Eurobox holding up to 126 locking cylinders, cylinder cores and other galvanized precision parts in a single pass. Four lighting scenarios generate four images per component, and AI inference then runs in about 10 milliseconds per cylinder. Engineers inspecting mirror-finish or galvanized parts will recognize the underlying problem: a single fixed lighting angle hides exactly the defects that matter, and manually rotating parts to catch every angle does not scale on a production line.
"Especially with highly reflective surfaces, it is essential to depict fine differences stably and in detail," said Jürgen Hejna, product owner for 2D cameras at IDS. "The camera's high sensor resolution and easy integration provide a solid foundation for industrial applications." Danube Dynamics said integrating the camera required no custom interface work, which cut development time and helped keep results stable once the system moved into ongoing production.
In its first live deployment, the system flagged deviations that experienced inspectors had missed, according to Danube Dynamics. Quality control is now fully digitized: results are logged and statistically evaluated automatically, and scrap and deviations that went unrecorded under manual inspection are now traceable.
EVVA can retrain the AI system on its own to add new part variants and defect patterns, according to Danube Dynamics, without altering the underlying camera and lighting setup. That self-service retraining loop is what the companies point to as the system's answer to a shop floor where locking-cylinder designs and finishes keep changing.
