Surface
Scratches, dents, contamination, coating gaps.
Application · Quality
Defects caught in real time, not in the returns bin.
Surface defects, assembly verification and print quality on the line. Optics and lighting first; classical rules when they hold; AI when variation wins.
What “good” must mean on your line
We translate vague quality complaints into measurable features: scratch vs acceptable texture, missing clip vs shadow, print smear vs contrast loss, then pick the stack that can see them at takt.
Assembly
Presence, orientation, correct part, clip engagement.
Print / label
Legibility, position, completeness.
Decision framework
classical Computer Vision vs AI on defects
Classical wins when defects are crisp and stable. AI earns its place when appearance varies too much for fixed rules, still on the line, not in a cloud demo.
| Defect character | Usual approach | Note |
|---|---|---|
| Known, geometric defects | Classical tools | Fast, explainable |
| Cosmetic / variable texture | AI classification / anomaly | Needs good samples |
| Mixed station | Classical + AI | Rules for hard facts, AI for greys |
Integration · Feasibility · Delivery
Integration, feasibility and delivery
Result to PLC or reject; PoC on good and bad samples; delivery from components to a full-service station.
Integration
Pass/fail, logging and reject timing coupled to your machine control.
Feasibility
Optics and lighting first, AI only when classical tools demonstrably fall short.
Delivery
Components & advice, integration, or complete handover via How we work.
Bring good and bad samples
We will say what is optically possible at your takt, and which architecture fits.