Feature Selection Based on Information Entropy for Accurate Detection of Optical Fiber End-Face Defects.

Area of Science:

• Optical Engineering
• Materials Science
• Image Processing

Background:

• Multimode fibers (MMF) are crucial for short-distance, high-bandwidth optical communication.
• Manual inspection of MMF end-face defects is time-consuming and error-prone.
• Existing automated methods struggle with impurities and lack accuracy.

Purpose of the Study:

• To develop an efficient and accurate automated method for detecting multimode fiber end-face defects.
• To overcome limitations of manual inspection and existing machine vision techniques.
• To enhance the reliability of optical fiber communication systems.

Main Methods:

• Proposed the WGMOS digital image detection method.
• Utilized information entropy for feature selection and optimization.
• Optimized image acquisition, correction, filtering, adaptive segmentation, and feature extraction.

Main Results:

• The WGMOS method effectively suppresses interference by minimizing background noise entropy.
• Achieved a significant increase in image equalization value (≥38.20%) compared to POL methods.
• Demonstrated a substantial improvement in signal-to-noise ratio (≥6.0%).

Conclusions:

• The WGMOS method offers superior impurity exclusion and detection accuracy for MMF end-face defects.
• This approach enhances efficiency and reliability in optical fiber inspection.
• Information entropy-based feature selection is key to improved automated defect detection.