The fragment count in the standardized fragmentation test defined in the EN 12150-1 standard is the way to define the safety level of tempered glass. Furthermore, it is also a way to get an indication about the stress and strength level of the tempered glass. Even though the way to count the number of fragments is defined in the standard by example, the actual result always depends on the examiner. To make the counting consistent across the examinations and examiners, an automated process of fragment counting is needed. This makes it a perfect application for a computer vision system since a computer never tires or loses its objectiveness. Systems for automated fragment counting already exist, but they have severe limitations.
Modern tools and technologies have revolutionized the field of computer vision in the recent years. Mainly this is due to the advances in convolutional neural networks, which are especially suitable for extracting patterns and information from visual imagery. To make fragment counting systems more flexible, faster and cheaper, these advances in computer vision can be utilized. Modern technologies allow an automated computer vision system to be implemented even on a mobile smartphone.