This system was implemented based on the Cognex D900 vision system with the ViDi OCR framework. Inscriptions (up to 10 lines) are engraved on the metal surface around the circumference of the cylindrical surface. Additionally, there is a stamped DataMatrix code. The system analyzes each line of text in terms of the correctness of the subtitles and their readability, and also analyzes the DataMatrix code (content and quality of the code). All non-conforming items are marked as non-compliant with quality requirements. Due to the multitude of references (several dozen different programs), OCR tools have been automated and thanks to this, the process of training neural networks is much faster than for standard OCR systems. The design was copied to a second analogous production line.