Implementation of a Programming Environment for Recognition of Metallurgical Production Objects in Real Time

The article is devoted to the creation and testing of a programming environment for an automatic identification system for steel sheets in metallurgical production based on the analysis of unique biometric features. The proposed solution ensures detection of a representative surface area and its subsequent recognition as the object moves along the process line. The proposed approach eliminates dependence on external markers, such as QR codes or tags, due to a detailed study of the surface texture, defect distribution and rolling patterns. To optimize the neural network model and simulate real operating conditions, a production line model was used that takes into account variable parameters such as lighting, object movement and image noise level. An important aspect of the study is a solution that allows performing recognition functions in real time and provides end-to-end integration of the system with adjacent industrial MES and SCADA platforms via the REST API.