Data augmentation using generative models for track intrusion detection.

Area of Science:

• Artificial Intelligence
• Computer Vision
• Railway Engineering

Background:

• Unauthorized railway track access poses a severe collision risk.
• Existing intrusion detection algorithms struggle with limited data and class imbalance.

Purpose of the Study:

• To develop a deep learning algorithm for detecting railway track intruders.
• To address data scarcity and imbalance issues in intrusion detection.

Main Methods:

• A hybrid algorithm combining generative models and classification networks was proposed.
• Generative models synthesized realistic intrusion data to augment limited datasets.
• Deep neural networks were trained on augmented data for intrusion identification.

Main Results:

• The algorithm effectively overcame limitations of scarce and imbalanced learning data.
• Augmented data using generative models led to improved intrusion detection accuracy.
• Evaluations on real datasets confirmed the algorithm's practical effectiveness.

Conclusions:

• The proposed algorithm offers a robust solution for railway track intrusion detection.
• Generative models enhance deep learning performance in safety-critical applications.
• This research underscores the potential of AI in improving railway safety systems.