Verfahren zur Identifizierung von Defekten in einer Verpackung

Details

Veröffentlichungs-Nr.

EP4752530

Anmeldetag

17. Oktober 2025

Veröffentlichung

3. Juni 2026

Rechtsraum

EP

IPC

G01N19/04, G06T7/00

Abstract

A computer-implemented method (800) for identifying defects in a package (102a,b) is disclosed. The method (800) comprises placing (802) the package in an apparatus for delaminating the package (102a,b), attaching (804) the first side (302a) to a first clamping device (402a) of the apparatus (400) and the second side (302b) to a second clamping device (402b) of the apparatus (400), tearing (806) apart the weld (104a,b,106a,b) by moving the first clamping device (402a) and second clamping device (402b) away from each other, obtaining (808) a first image (600b), depicting the weld (104a,b, 106a,b), at a first point of time (t1), obtaining (810) a second image (600c), depicting the weld (104a,b,106a,b), at a second point of time (t2), wherein the second point of time (t2) is subsequent to the first point of time (t1), processing (812) the first image (600b) by identifying a sealing line (404)at the first point of time, processing (814) the second image (600c) by identifying the sealing line (404) between the first and second side (302a,b) of the laminate at the second point of time, wherein the first and second image (600b,600c) input to the machine learning model (414) comprise the sealing line (404) identified, applying (816) a machine learning model (414), wherein the machine learning model receives as inputs the first and second image (600b,c), and outputs a prediction of the defects, wherein the machine learning model (414) has been trained by means of reference first and second reference images indicative of the defects for reference packages associated with the first and second reference images.

Anmelder

Firma

Tetra Laval Holdings & Finance S.A.

Land

🇨🇭 Schweiz