Nonlinear and Dotted Defect Detection with CNN for Multi-Vision-Based Mask Inspection.

Area of Science:

• Quality Control
• Computer Vision
• Deep Learning

Background:

• Increased global mask production necessitates more efficient defect detection systems.
• Traditional computer vision methods struggle with small, varied nonlinear and dotted mask defects.
• Existing mask inspection systems require enhanced algorithms for improved accuracy.

Purpose of the Study:

• To develop an advanced deep learning-based defect detection method for mask manufacturing.
• To address the limitations of traditional algorithms in identifying subtle mask defects.
• To create a robust system applicable to real-world manufacturing environments.

Main Methods:

• Implementation of a convolutional neural network (CNN) for defect identification.
• Integration of efficient preprocessing techniques tailored for mask inspection data.
• Training and validation using real-world data from actual mask manufacturing lines.

Main Results:

• Successful detection of nonlinear and dotted defects on masks.
• Demonstrated superior performance compared to previous defect detection methods.
• Validation of the method's efficacy in a production setting.

Conclusions:

• The proposed deep learning method effectively detects critical mask defects.
• This approach offers a significant improvement over traditional computer vision techniques for mask quality control.
• The system is viable for real-time application in mask manufacturing lines.