Integrating Deep Learning Based Anomaly Detection with Extended Reality: A Case Study on Extensive Railways Monitoring
Maria di Summa, Angelo Cardellicchio, Nicola Mosca, Marina Ricci, Vito Renò, Udith Krishnan Vadakkum Vadukkal, Ettore Stella · 2024
The highly demanding safety standards adopted in the railway context imply that cutting-edge technologies must be used to limit accidents and ensure their complete avoidance. As such, developing integrated monitoring systems, which also exploit eXtended Reality technologies along with deep-learning-based anomaly detection techniques, becomes crucial to support the awareness of a planning operator throughout the maintenance operations required to comply with high-quality standards. This work addresses the abovementioned problem by proposing a framework composed of three different steps: data collection and preparation, anomaly detection via deep neural networks, and presentation of the achieved results. Specifically, the final step involves displaying the anomaly detector results in a virtual environment, reproducing the railway line under analysis. This environment will provide the planning operator with a complete platform to explore, use to plan maintenance interventions, and gather detailed reports to improve the overall safety of the railway line effectively.