Deep learning-based identification of steel products

To achieve smart manufacturing in Industry 4.0 for steel industry (or Steel 4.0), this paper proposes a smart steel manufacturing framework, where a deep learning-based automatic identification tracking method for steel products is developed. Automatically online tracking and identifying steel products on a production line is essential for smart manufacturing management since those products might be frequently moved and processed everywhere on the product flow. Existing approaches usually rely upon marking or embedding a series of identification codes on the steel surfaces. However, steel-making is usually processed under a very high temperature environment, making it difficult to well embed the identification codes with acceptable quality for further automatically online recognizing them. To tackle this problem, this paper presents a vision-based automatic identification tracking method without needing to embed any identification codes onto the steel product surfaces. The key idea is to utilize the essential identity of a steel product without extrinsic information embedded, achieved by automatically and deeply learning visual features from the steel image. The presented preliminary results have verified the efficiency of the proposed method.