Multimodal Displacement Detection for Digital-Twin Synchronization of Underground Utility Tunnels in Smart Cities

A digital twin (DT) of an underground utility tunnel is valuable only if it remains synchronized with the physical asset. The objective of this study is to maintain this synchronization for an operational tunnel by automatically detecting displacements of linear and point assets, using only the fixed sensors already deployed in the tunnel. However, most DT research emphasizes design and construction, rather than operational-phase change detection. The BIM–GIS-based framework of our previous study was limited to relative-coordinate reporting, offline updating, and indoor mock-up validation. To this end, this study introduces a multimodal displacement detection framework with an automated, event-driven update workflow, which addresses these limitations and is demonstrated in an operational tunnel. Linear-object sag (power cables) is extracted from sparse, fixed fiber-scan LiDAR by adapting a recursive terrain-fragmentation ground filter into a wall-as-ground context. Point-object displacement (fire extinguishers) is detected using a single fixed CCTV camera, combining a YOLOv5 detector with a perspective transform to recover absolute coordinates in a GNSS-denied corridor and differentiate between moved and toppled assets. Detected displacements are serialized as JSON and streamed via a Kafka broker into the DT service environment, integrating heterogeneous events into a unified synchronization workflow. Field validation against a co-located survey-grade scanner confirmed agreement within the 0.10 m operational validation tolerance for nominal sags of 7, 10, and 15 cm, with the largest residual at the smallest deflection. The framework maintains the virtual model consistent with a physical tunnel without additional instrumentation, enhancing disaster prevention and response.