- El-Cezeri
- Vol: 9 Issue: 4 Özel Sayı
- Open Package Detection with Deep Learning Algorithms Using Thermal Camera in Heat Sealed Packages
Open Package Detection with Deep Learning Algorithms Using Thermal Camera in Heat Sealed Packages
Authors : Engin Karataş
Pages : 1363-1374
Doi:10.31202/ecjse.1135411
View : 10 | Download : 3
Publication Date : 2022-12-31
Article Type : Research
Abstract :In the food industry, packaging is as important as the product inside, for reasons such as protecting the product in the package, ensuring the safety of its content in terms of health, and gaining the appeal of the consumer. Packages that are not properly closed cannot protect the product from external factors such as humidity and temperature, and they also pose a risk to health as they will cause the product to deteriorate earlier than expected. Packages that remain open reduce the consumer\'s perception of brand and product quality and cause customer complaints. There are different technologies to control proper closing and sealing of the packages. Some of these are vacuum-based systems, ultrasonic control systems, machine vision systems using X-ray and camera images. In this study, open package detection in heat-sealed packages with transparent packaging is studied. The defects in the transparent packaging could not be seen in the images taken with standard industrial cameras. The shape of the packaging machine jaw temperature on the package where the jaw presses can be seen with a thermal camera. This shape, similar to the letter \'T\', was tried to be classified in open and closed packages by using blob analysis, geometric matching and support vector machine in preliminary studies, but low success rates were obtained. In order to achieve a high success rate, firstly \'Convolutional Neural Network Models\' were tried and an accuracy rate of around 95% was obtained. Then, MobileNet and ResNet networks were used with the \'Transfer Learning\' method to increase the success rate, and a success rate of over 99% was achieved.Keywords : mühür denetleme, evrişimsel sinir ağı, öğrenme aktarımı, termal kamera