Centroid growth selective clustering method for surface defect detection in silicon nitride ceramic bearing rollers

Area of Science:

• Materials Science
• Mechanical Engineering
• Computer Vision

Background:

• Surface defects on silicon nitride ceramic bearing rollers present significant image segmentation challenges due to fuzzy edges and gradient plunges.
• Existing methods struggle with contour anomalies, incomplete segmentation, and notch misidentification for these defects.

Purpose of the Study:

• To propose and validate the Centroid Growth Selective Clustering Method for accurate detection and segmentation of fuzzy surface defects.
• To enhance the reliability and performance of silicon nitride ceramic bearing rollers in high-precision applications.

Main Methods:

• Analyzing discontinuities in notch regions of fuzzy edges to determine image centroid via Euclidean distance probabilities.
• Applying hierarchical clustering for effective image content separation and feature connectivity detection.
• Ensuring high-precision overlay and extraction of target surface defect features.

Main Results:

• Achieved 97.85% segmentation accuracy for fuzzy surface defects.
• Reached 95.71% edge coverage rate and over 85% intersection and concurrency ratios.
• Demonstrated significant mitigation of fuzzy edge impacts, improving defect detection and extraction.

Conclusions:

• The Centroid Growth Selective Clustering Method effectively addresses challenges in segmenting fuzzy surface defects on silicon nitride ceramic rollers.
• The proposed method enhances defect detection accuracy and enables complete surface defect extraction.
• Improved defect detection contributes to enhanced reliability and performance of ceramic bearing rollers.