Real-time detection method for bulk bubbles in optics based on deep learning

Area of Science:

• Optical Engineering
• Materials Science
• Artificial Intelligence

Background:

• Bulk bubbles in optical components can degrade laser-induced damage thresholds and beam quality.
• Manual inspection methods for bubbles are imprecise and inconsistent, hindering quality assurance.
• Automated detection is crucial for reliable optical manufacturing and performance.

Purpose of the Study:

• To develop a real-time, deep learning-based method for detecting bubbles within optical materials.
• To enhance the precision and consistency of bubble detection compared to manual techniques.
• To provide accurate measurements of bubble positions and radii for quality control.

Main Methods:

• A deep learning model was trained for real-time bubble detection in optics.
• The method was evaluated for its speed (frames per second) and detection accuracy (recall).
• Bubble radius retrieval accuracy and computational time were also assessed.

Main Results:

• The deep learning method achieved real-time bubble detection at 67 frames per second.
• A recall of 0.836 indicates high accuracy in detecting bubbles.
• Average absolute deviation for radius retrieval was 3.73%, with a processing time of 58.8 ms per bubble.

Conclusions:

• The proposed deep learning method offers real-time and accurate detection of bubble positions and radii in optics.
• This technology has significant potential to improve quality assurance in optical manufacturing.
• Automated, AI-driven inspection can replace manual methods, leading to more reliable optical components.