Battery screen print defect detection based on stationary velocity fields neural network matching and optical flow...

Area of Science:

• Manufacturing Engineering
• Computer Vision
• Materials Science

Background:

• Screen printing is crucial for battery manufacturing.
• Defect detection is vital for quality control.
• Existing methods struggle with irregular distortions.

Purpose of the Study:

• To develop an automated defect detection method for screen printing in battery manufacturing.
• To address challenges posed by irregular shape distortion.
• To improve detection accuracy and efficiency.

Main Methods:

• Utilized stationary velocity field (SVF) neural network template matching.
• Employed the Lucas-Kanade (L-K) optical flow algorithm.
• Implemented image preprocessing, template creation, image registration, distortion rectification, and defect extraction.

Main Results:

• Achieved high accuracy (97%) and time efficiency (485 ms).
• Demonstrated effective detection of defects like lacking, skew, and blur.
• Successfully handled irregular print distortions.

Conclusions:

• The proposed method offers superior defect detection for battery screen printing.
• It provides advantages in image registration, defect extraction, and industrial efficiency.
• The dynamic template updating mechanism enhances environmental adaptability.