Shannon Entropy Used for Feature Extractions of Optical Patterns in the Context of Structural Health Monitoring

Area of Science:

• Signal Processing
• Information Theory
• Machine Learning

Background:

• Technical vision systems (TVS) face challenges with optoelectrical signal noise during data acquisition.
• Electrical fluctuations and optical noise impact the accuracy of metrology subsystems.

Purpose of the Study:

• To propose a novel signal processing method for TVS using Information Theory (IT).
• To enhance the accuracy of optoelectronic signal classifiers for structural health monitoring (SHM).
• To improve spatial coordinate measurement performance in noisy environments.

Main Methods:

• Feature extraction of optical patterns using Linear Predictive Coding (LPC) and Autocorrelation Function (ACC).
• Application of Shannon Entropy segmentation (SH) for relevant information extraction.
• Classification of optical patterns using five different machine learning (ML) techniques.

Main Results:

• Shannon Entropy segmentation achieved classification accuracy exceeding 95.33%.
• Performance without Shannon Entropy segmentation resulted in a worst-case accuracy of 33.33%.

Conclusions:

• Shannon Entropy segmentation significantly enhances the accuracy of optoelectronic signal classifiers in TVS.
• The proposed method improves structural displacement estimation and health evaluation under noisy conditions.