In industrial production, product serial numbers are often designed to be resistant to environmental influences such as emulsions, metal shavings, paint, oil, and similar contaminants.
Due to regulatory requirements, these markings are made to be durable and easily readable by humans. However, they are not optimized for automated reading in automatic identification systems, which often leads to redundant labeling and increased operational costs. This thesis addresses the development of a machine vision system capable of automated reading of product serial numbers using smart cameras. The fundamental concepts of traceability, optical character recognition, image processing, and the operation of smart cameras are first introduced. A prototype system is then developed, comprising a camera, appropriate illumination, test samples, and a software solution with a user interface. The system is evaluated through image acquisition, parameter configuration for recognition, and training of the reading algorithm. Based on experimental results, the recognition accuracy, system limitations, and potential improvements are analyzed.
|
