The continuous casting surface quality inspection system comprehensively applies multi-disciplinary knowledge such as light, mechanics, electricity, pattern recognition, and deep learning. It is a typical application of machine vision AI recognition technology in the metallurgical industry.
1) Seize the “export gate” of steelmaking and improve the quality control level of the process
2) Reduce on-site personnel, reduce labor load, and increase the effect of centralized control of continuous casting operations
3) Integrate with the production control system (L1, L3) to automatically control the continuous casting billet logistics rhythm according to the detection results
4) Increase the hot transfer rate of continuous casting slabs
The optical imaging system ensures that the detection device can obtain high-contrast defect images with optimal resolution, optical path and imaging method, ensuring that defects are most clearly visible in the background, laying a solid foundation for subsequent image processing and defect classification.
The area array and line cameras of the optical imaging system are designed with infrared cut-off filters, LED high-brightness light sources, strobe light sources, custom-selected large aperture lenses and other imaging devices to adapt to the surface conditions of continuous casting and overcome high temperatures and red heat. interference with imaging.
1) Intuitive display and complete data storage
2) As a supplement to automatic defect identification, it prevents serious defects from being missed.
3) Sensors can be added on both sides of the end for end-face quality detection
4) Facilitates defect collection and can be marked manually
The area array camera module, combined with line array data, completes defect detection and classification based on AI learning in the image processing computer.
The system can be configured with an automatic recognition module for the continuous casting billet number, and uses an external triggering method to shoot, that is, the position of the continuous casting billet is detected through the through-beam switch. When the billet moves to cover the through-beam switch, the light source strobes, and the camera captures an image simultaneously. This image collection method can ensure that the relative position of the character surface of the continuous casting billet and the camera is fixed at the moment of collecting the image. The billet surface is located at the best focus position of the camera, and the image has the best clarity and resolution lock.
Based on the deep learning algorithm, character recognition can achieve higher accuracy and robustness, and compare and match the recognized characters with L2 information. For unrecognizable characters, an alarm will be given.
Save labor, improve working environment, improve efficiency, reduce error rate and save cost！
Сэкономьте труд, улучшите рабочую среду, повысьте эффективность, уменьшите количество ошибок и снизьте затраты！