Laser Marked Code Reading
to Catch Unreadable Marks at the Source

Laser marked code reading on metal is difficult because the mark is often only slightly darker or rougher than the surrounding surface. Reflective metal adds glare that can hide it completely. Curved or textured parts make the problem worse. Low-angle lighting, coloured light or polarisation can lift the mark, and AI models trained on your faint examples read what remains reliably. The same station can then confirm every mark before the part moves on.

Yes. Many parts carry a Data Matrix code alongside human-readable text, such as a serial or part number. KOMPASS reads both from the same image and can check that they agree, which catches marking errors where the code and the printed characters were generated from different data. That check is hard to do by eye on a fast line. It also confirms that the serial follows the expected sequence.

No. Reading decodes the mark and checks its content. Formal grading of mark quality against ISO/IEC 29158 uses a dedicated verifier with controlled optics and is done where your customer or standard requires it. KOMPASS focuses on whether each mark can be read and carries the right data, and it trends reading confidence to show when marks start to degrade.

2D codes include error correction, so they can still decode with some damage. That is useful, but it can hide a marker that is slowly getting worse. By tracking how confident each read is, not only whether it decoded, the station can flag marks that are heading toward failure before they stop reading altogether downstream. That gives maintenance a reason to clean a lens or check focus before parts are lost.

At or just after the marking station is usually best. A bad mark found there can be held or re-marked before the part enters assembly or heat treatment. Placing the camera beside the marker also means the expected data is available immediately for comparison. Later checkpoints are useful when coatings or processes may affect the mark. The marker and camera can also share a trigger, so each mark is captured as soon as it is finished.

The same approach applies to other direct part marking methods, such as dot peen and chemical etching, because each relies on surface change rather than ink. The model is trained on the methods you use. DPM code reading on cast or rough surfaces may need different lighting from laser marks on machined faces, which is settled in the trial.

Each reading is stored with the decoded data, a time stamp and the image, creating a traceability record for every part from the moment it was marked. If a part fails later, the record shows how its mark looked when it was made. That helps settle disputes about whether a mark was damaged in transit or marked poorly. It also gives quality teams a simple way to answer customer questions about a specific serial.

Send marked parts, including faint and damaged examples, the code types and text you mark, and how marking data is generated. Tell us the marking cycle time and where a failed part should go. With that, we can assess lighting, camera position and integration, and plan monitoring at the marker. Parts marked across several shifts help show how the marker drifts over time.