AI-based Monitoring of Urban Public Transport Safety Using Computer Vision
Keywords:
computer vision, public transport safety, anomaly detection, real-time monitoring, edge computing, privacy-preserving AIAbstract
Urban public transport systems are essential for city mobility but face persistent safety challenges—vandalism, overcrowding, slip-and-fall incidents, theft, and near-miss collisions at stops. This paper presents a comprehensive framework for real-time monitoring of urban public transport safety using computer vision and deep learning. The proposed system integrates multi-camera feeds, edge-compute modules, and a lightweight deep neural pipeline for event detection (falls, fights, crowding, unattended objects), anomaly scoring, and operator alerting. We evaluate the approach on a mixed dataset collected from bus interiors, tram platforms, and bus stops, achieving an average precision of 0.88 for event detection and a mean time-to-alert under 2.2 seconds on edge hardware. We also discuss privacy-preserving strategies, deployment considerations, and a roadmap for integrating the approach with existing transport management centers.
