While robot-based inspections show promise in the transportation infrastructure context, expert visual inspections are still essential and mandated for assessing asset safety. However, these processes remain heavily reliant on traditional manual methods, causing significant information losses and inefficiencies. To support such fragmented activities, this paper introduces an automatic registration (geometric and semantic alignment) system for Augmented Reality (AR), facilitating data and information exchange between the field and a Digital Twin (DT) platform for infrastructure management that exploits Linked Data technologies. High-accuracy 6-DoF pose estimation of the AR interface (e.g., Microsoft HoloLens2) in large scale, open unprepared environments is achieved by using a hybrid approach based on Real-Time Kinematic (RTK) and Visual Inertial Odometry (VIO). To this end, an in-house developed RTK receiver is rigidly attached to the AR visor. This system automatically provides field operators with on-field access and visualization of existing geo-registered DT information (e.g., BIM models) via AR and facilitates the enrichment of the asset’s DT model with newly captured data (e.g., images). This enables large scale information registration, linking, accessing and querying, improving the efficiency of civil infrastructure inspections. The proposed system is tested on a real use case, consisting in a routine visual inspection of a highway viaduct.