Monocular line scan vision-based surface defect detection approach for highly reflective bearing balls

Area of Science:

• Mechanical Engineering
• Optical Engineering
• Materials Science

Background:

• Surface defects significantly impact the stability and lifespan of bearing balls.
• Existing detection methods may struggle with the unique challenges of bearing ball surfaces, such as curvature and reflectivity.

Purpose of the Study:

• To develop a monocular line scan vision-based detection system for identifying surface defects on bearing balls.
• To address challenges related to nondevelopability, spherical curvature, and high reflectivity of bearing ball surfaces.

Main Methods:

• Designed a specialized optical system and illumination principle for bearing balls.
• Developed a line scanning unfolding process and image acquisition scheme based on motion unfolding trajectory analysis.
• Established a mathematical model for whole-surface unfolding and created a simulation.
• Implemented a spatiotemporal image defect detection algorithm using subtraction and spatial-temporal resolution normalization.

Main Results:

• Achieved a detection resolution of approximately 0.001 mm² for crack defects.
• Demonstrated a 100% detection rate for crack defects.
• Validated the high accuracy of the proposed defect detection method.

Conclusions:

• The developed vision-based system effectively detects surface defects on bearing balls.
• The proposed method offers high accuracy and resolution for critical defect identification.
• This technology enhances the quality control and extends the service life of bearing balls.