Strain FBG-Based Sensor for Detecting Fence Intruders Using Machine Learning and Adaptive Thresholding.

Area of Science:

• Engineering
• Computer Science
• Physics

Background:

• Intruder detection systems are crucial for security.
• Optical fiber sensors offer a promising technology for perimeter security.
• Challenges exist in detecting intruders under low signal-to-noise ratio (OSNR) conditions.

Purpose of the Study:

• To develop and demonstrate an intruder detection system.
• To classify different scenarios including no intruder, intruder, and wind.
• To evaluate the system's performance at low OSNR levels.

Main Methods:

• Utilized a strain-based optical fiber Bragg grating (FBG) sensor.
• Implemented machine learning (ML) algorithms, specifically linear discriminant analysis (LDA) and logistic regression.
• Employed adaptive thresholding to enhance classification accuracy.

Main Results:

• The system successfully classified intruders, non-intruders, and wind disturbances.
• Adaptive thresholding significantly improved ML classifier performance in low OSNR environments.
• An average accuracy of 99.17% was achieved at OSNR levels below 0.5 dB.

Conclusions:

• The proposed FBG sensor-based system with ML and adaptive thresholding is effective for intruder detection.
• The method demonstrates robustness and high accuracy even in challenging low OSNR conditions.
• This approach offers a reliable solution for securing perimeters, such as college gardens.