SAFETRACK: INTELLIGENT ANOMALY DETECTION FOR RAILWAY STATION SAFETY THROUGH UNSUPERVISED LEARNING

Abstract

Railway stations are complex environments characterized by high human traffic, dynamic operations, and the potential for safety-critical incidents. Traditional monitoring and safety management systems rely heavily on manual supervision and rule-based algorithms, which often fail to detect unforeseen or evolving risk situations. To address these limitations, this paper introduces SafeTrack, an intelligent anomaly detection framework that leverages unsupervised machine learning to enhance safety management and accident prevention in railway stations. The proposed system utilizes a combination of sensor networks, surveillance footage, and operational data to automatically identify irregular patterns associated with unsafe behaviors or hazardous conditions. By employing clustering and autoencoder-based anomaly detection methods, SafeTrack can recognize deviations from normal operational behavior without the need for labeled datasets. This allows for continuous learning and adaptation to new safety threats, such as overcrowding, equipment malfunctions, or passenger distress, in real time. Extensive simulations and prototype evaluations demonstrate that SafeTrack significantly improves detection accuracy and reduces false alarms compared to conventional rule-based monitoring systems. Moreover, its scalable architecture enables seamless integration with existing railway management infrastructure, ensuring rapid deployment and low maintenance. The findings suggest that unsupervised learning approaches hold substantial potential for achieving autonomous, data-driven safety intelligence in transportation hubs. Ultimately, SafeTrack provides a step forward toward the realization of smart, selfaware railway environments capable of proactive risk mitigation and efficient emergency response.