Image processing based quality control of the impermeable seams in multilayered aseptic packages

• We propose an approach to detect unsteady seams in multilayered aseptic packages.
• It is the first study on quality control of impermeable seams.
• Segmentation and edge detection are evaluated on longitudinal and transverse seams.
• One feature extracted from images is classified by linear regression and SVM.
• The proposed system can overlook an unsteady packet only once in a 96 years.

Multilayered aseptic material which guarantees the extended shelf life of liquid foods is turned into an impermeable package by folding packaging machines. Some problems in the machines causes the packages to lose the impermeability property and therefore lead to deterioration of the liquid food. The leak test, which was performed to determine the problem, has been carried out by expert employees. The control process has steps of random selection of the liquid food packages, the opening of it properly, and observing the distribution of the injected ink in the seams by human eye. It is known that this control can cause serious material damage due to the human-based errors. With this study, it was aimed to perform the leak test of seams in multilayered aseptic packages by using a combination of the image processing and classification techniques. Through the data obtained from a real production environment, the experiments were performed and the results were evaluated. As a result, it can be said that the study has a distinctive feature as being the first in its field with its promising results.