| Popis: |
In the beverage manufacturing industry - the process of closing and sealing the bottle caps by an automated machine could cause defects due to imperfect equipment alignment. Small businesses rely on humans as a means to perform quality control, which is prone to errors due to long working hours. In this paper, we propose our real-time Deep Learning-based technique to automatically detect and classify bottle caps into 2 categories: good and defective. Good bottle caps should be aligned perfectly with bottle rings and close to the bottlenecks while defective bottle caps include loose caps, tilted caps, and missing caps. Our approach consists of combining Image Processing with Convolutional Neural Network. This machine learning approach is versatile, allowing the system to work even with different light and background settings. The system is suitable to perform at low to moderate conveyer speed capable of processing 24 frames per second with 96.40% overall accuracy. |
