PSI - Issue 78

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

1287

actual structure. However, structural damages are not directly observable in the point cloud due to its sparse nature. Therefore, it is essential to enrich the 3D model with aligned visual data to enable effective condition assessment. Specifically, the images of the cracks are collected through drone flights and matched to the 3D map via relocalization techniques, and finally displayed in the VR scene using interactive markers. When an operator selects one of these markers, a temporal sequence of corresponding images is shown, allowing users to observe the progression of damage over time and compare multiple inspections from the same viewpoint. This temporal anchoring of visual data in 3D space provides an effective mechanism for damage tracking and decision support. The five main stages of the proposed methodology are shown in Fig. 1 and outlined below.

Fig. 1. Overview of the proposed framework.

2.1. 3D Mapping The initial stage involves the acquisition of a 3D point cloud map of the structure using a LiDAR sensor. In this study, we employ the X120Go SLAM Laser Scanner, which performs continuous 3D mapping while the operator moves through the environment. The SLAM algorithm simultaneously estimates the sensor ’ s trajectory and reconstructs the surrounding geometry in real-time, allowing the generation of a coherent point cloud without the need for fixed scanning stations. The X120Go acquires up to 200,000 points per second with a spatial resolution of approximately 6 mm, and features a 360° horizontal and 270° vertical field of view. This ensures the generation of a unified and high-resolution 3D model of the entire environment. 2.2. Image Acquisition A Micro Aerial Vehicle (MAV) equipped with a ZED 2i stereo camera is employed to perform regular inspection flights, capturing high-resolution images. The use of an aerial platform offers significantly greater flexibility in navigating complex environments that are difficult or unsafe to access. During each inspection mission, the MAV follows a preplanned or partially autonomous trajectory, capturing stereo images along its path. These image sequences are employed for further analysis required in the successive steps. 2.3. Identification of Structural Cracks Videos collected by the MAV are currently reviewed by a domain expert, who inspects each sequence and annotates the frames in which visible structural cracks or defects are present. Despite the involvement of a human operator is