Small Defects Detection of Galvanized Strip Steel via Schatten-p Norm-Based Low-Rank Tensor Decomposition

Area of Science:

• Materials Science
• Computer Vision
• Image Processing

Background:

• Surface defect detection is vital for galvanized strip steel quality.
• Small targets in low-contrast images present a significant detection challenge.
• Existing methods struggle with the small pixel occupation and low contrast of white-spot defects.

Purpose of the Study:

• To develop an accurate and efficient method for detecting white-spot defects on galvanized strip steel.
• To address the limitations of current defect detection techniques for small, low-contrast targets.
• To improve the quality control in steel production through enhanced defect identification.

Main Methods:

• Proposed a novel Schatten p-norm-based low-rank tensor decomposition (SLRTD) method.
• Transformed defect images into a patch-based tensor mode for enhanced information mining.
• Incorporated a nonlinear reweighting strategy using the Schatten p-norm to improve decomposition performance and address over-shrinkage issues.
• Developed a solution framework using an alternating direction method of multipliers for defect image extraction.

Main Results:

• The SLRTD method successfully decomposes defect images into low-rank background, sparse defect, and random noise components.
• Experimental results on a real-world galvanized strip steel dataset demonstrate superior performance compared to state-of-the-art methods.
• The proposed method achieves both qualitative and quantitative improvements in white-spot defect detection.

Conclusions:

• The SLRTD method offers a robust and effective solution for detecting challenging white-spot defects in galvanized strip steel.
• This approach significantly enhances the accuracy and efficiency of surface defect detection in industrial quality control.
• The findings contribute to advancing automated inspection systems in steel production.