Automated surface inspection for steel products using computer vision approach

Area of Science:

• Materials Science
• Computer Vision
• Industrial Automation

Background:

• Automated surface inspection is crucial for steel product quality.
• Existing computer vision methods struggle with non-uniform illumination and similar surface textures/defects.
• Limitations hinder widespread adoption in diverse inspection scenarios.

Purpose of the Study:

• To develop a robust framework for automated steel surface inspection.
• To overcome limitations of current methods related to illumination and texture variations.
• To enhance the consistency and efficiency of defect detection in steel manufacturing.

Main Methods:

• Investigated the image formation process to model illumination's impact on image quality.
• Developed a quantitative model to remove non-uniform brightness.
• Designed a simple classifier to distinguish defects from surface textures.

Main Results:

• The proposed framework demonstrates robustness to illumination changes.
• Achieved successful real-time surface inspection of round billets in manufacturing.
• The algorithm processes at 12.5 frames per second with over 90% detection rate for turned and skinned billets.

Conclusions:

• The novel framework offers a significant improvement for automated steel surface inspection.
• Its robustness and wide applicability make it suitable for various inspection scenarios.
• The approach enhances product quality control in the steel industry through reliable defect detection.