PSI - Issue 78

Simone Felicioni et al. / Procedia Structural Integrity 78 (2026) 1285–1292

1290

4.2. Data processing

The proposed framework is designed to be general and modular, allowing for both manual and automated data analysis workflows. In the current implementation, damage identification is performed manually by an expert operator, who inspects the video stream collected by the MAV and selects the frames containing structural cracks or defects. However, this process is inherently extendable to an automated pipeline. In future implementations, an object detection model trained for damage recognition could be employed to automatically identify and extract frames of interest from the acquired video. This would remove the dependency on human supervision and improve scalability, allowing for the monitoring of larger areas and more frequent assessments over time. It is worth noting that the localization of the detected damage in the 3D model is already automated. Once a frame containing damage is selected, a visual relocalization module matches it with the corresponding area in the 3D map as shown in Fig. 4, allowing the projection of POIs directly onto the digital replica.

Fig. 4: Example of (a) a 3D-to-2D projection of the point cloud and (b) an image collected by the MAV from a similar viewpoint.

4.3. VR headset The Meta Quest Pro has been employed in this study as HDM for the immersive visualization of the digital replica. Fig. 5 depicts the headset, while Table 1 reports its specifications. The HDM allows users to move within the virtual reconstruction of the structure and interact with the elements of interest. Interaction is enabled through both handheld controllers and hand-tracking capabilities, allowing users to select and activate the markers associated with the POIs.

Table 1. VR headset specifications.

Display

Resolution

Memory

2 x QLED

1800x1920

12 GB

Field of View

Sensors

Connectivity

106° horizontal 95.57° diagonal

Eye, face, hand, body tracking Ambient light Accelerometer Proximity sensor Gyrometer Barometer Magnetometer

WiFi 6E Bluetooth 5.2 USB – Type C

Human Interaction

Audio

Battery Life

Controllers Hand gestures

Speaker Microphone

2 hours

CPU

Storage

Weight

Octa-core Kryo 585

256 GB

722 g