Eine simulierte Zeilenkamera schaut auf eine laufende Folienbahn und findet Gels, Kratzer und Löcher. Sie können Bahngeschwindigkeit und Fehlerdichte verändern und sehen, wie Erkennung, Verfolgung und Qualitätssiegel reagieren. Achten Sie auf die Rolle des MES-Kontexts: Erst mit Maschinendaten wird aus Bildern eine zertifizierbare Rolle.
A simulated line camera watches a running film web and finds gels, scratches and holes. Change line speed and defect density and watch detection, tracking and the quality seal react. Note the role of MES context: only with machine data do images become a certifiable roll.
Calender / extruder line · simulated line camera · defect detection, tracking and roll certification in real time
| Time | Class | Confidence | MD position | CD position | Size | Area |
|---|
A synthetic backlit camera image of a running film web is generated frame by frame (texture, uneven illumination, defects injected with
material-dependent probabilities). The detector estimates the local background with a 31 × 31 box filter, thresholds the deviation
(|pixel − background| > 14), bridges thin gaps vertically and labels connected components. Shape decides the class:
saturated pixels → hole (backlight through the film), aspect ratio ≥ 3 along machine direction → scratch, otherwise gel.
Each detection carries a confidence derived from contrast and shape margin. A tracker follows every defect across frames using the known web advance,
so each is counted once with its MD/CD position and geometry. Defects per 100 m against the roll limit drive the certification verdict.
Everything you see is simulated and computed locally in your browser — no camera, no server, no customer data. Order and customer names are fictional. The desktop version of this system (Python · OpenCV · PyQt6) adds a line-camera interface, roll maps and MES/OPC UA connectors.