Technical Workflow Overview
Following the initial contact, a series of meetings were organised to discuss the aims of the research project, clarify the research questions, and understand the archaeological and environmental contexts in which we would operate. Because the site is located on Gotland, all preparatory discussions and planning sessions were carried out online. During these meetings we defined the acquisition strategy, the types of data required, and the instruments necessary to support the research objectives.
The fieldwork was conducted through a joint effort between the School of Aviation and DARKLab. The School of Aviation deployed drones equipped with RGB and LiDAR sensors to acquire a high-resolution multisource model of the wider landscape, ensuring detailed morphological information of both the terrain and the monument, including the medieval tower of Sundre. In parallel, the tower was documented internally and externally using a FARO phase-shift laser scanner. A complementary photogrammetric acquisition was carried out to record high-quality colour information. The full acquisition campaign required three days: two days on site and one day dedicated to travel.
After fieldwork, the processing pipeline began by aligning and cleaning the laser-scanning data. The registered scans were then imported into Reality Scan for integration with the image-based datasets from the photogrammetry campaign, as well as with the LiDAR-derived landscape model. This produced a unified, high-resolution 3D dataset capturing both the monument and its surrounding environment at different scales. The LiDAR data were additionally processed to generate a large-scale morphological model suitable for extended landscape analysis.
The combined datasets were subsequently analysed using GIS for terrain and morphological interpretation. An AI-assisted analysis of the tower’s masonry was performed using TagLab (https://taglab.isti.cnr.it/), enabling automated annotation and segmentation of wall-surface features. All raw datasets, intermediate products, and final processed models were delivered to the project. To facilitate broad access, the 3D material was also published through 3DHOP (https://3dhop.net/), allowing researchers without specialised software or advanced 3D skills to interact with the models directly via a web browser.
Drone datasets were post-processed in CloudCompare, while the laser-scanning data underwent initial cleaning and refinement in MeshLab. The final integrated outputs were prepared according to the project’s requirements and made fully accessible for further research and interpretation.
1-LiDAR Acquisition – School of Aviation
The landscape surrounding Sundre was recorded using professional-grade UAV platforms operated by the School of Aviation. The drones were equipped with RGB cameras and airborne LiDAR sensors, allowing the acquisition of both photogrammetric imagery and high-density point clouds. The LiDAR system produced a detailed morphological model of the terrain, vegetation, and built structures, enabling accurate elevation modelling and feature detection at landscape scale. These datasets formed the foundation for building a multisource model of the entire area, ensuring that the environment of the medieval tower could be analysed with centimetric precision.
4-Data Integration – RealityCapture
All geometric datasets—including laser scans, photogrammetry, and landscape LiDAR—were integrated and processed in RealityCapture, a software optimised for hybrid workflows. The FARO scans were first registered and cleaned before being imported as structured laser data. High-overlap images of the tower and landscape were aligned using feature-matching, and the resulting models were merged with the aerial LiDAR dataset. This produced a unified 3D model covering multiple scales, from the fine detail of the tower’s masonry to the wider terrain morphology. The integrated mesh was then cleaned, decimated, and textured for further analysis.
7-AI-Assisted Analysis – TagLab
The detailed 3D model of the tower’s exterior was further analysed using TagLab, an AI-assisted annotation platform developed at CNR ISTI (https://taglab.isti.cnr.it/). TagLab enabled automated segmentation and classification of masonry units, facilitating a systematic analysis of stone types, dimensions, and surface conditions. The tool’s semi-automatic labelling functionality allowed researchers to rapidly annotate large surfaces while maintaining expert-level precision, producing datasets that support architectural interpretation and conservation planning.
2-Phase-Shift Laser Scanning – FARO Focus
The medieval tower of Sundre was documented inside and outside using a FARO Focus phase-shift 3D laser scanner, capable of capturing highly accurate geometry with minimal noise. Multiple stations were positioned around and within the tower to ensure full coverage of walls, openings, and architectural irregularities. The resulting scans provided sub-centimetre resolution geometry essential for structural interpretation and subsequent AI-driven masonry analysis. The scanner’s colour-capture mode was disabled in favour of dedicated photogrammetric imagery, maintaining optimal geometric quality while avoiding mixed datasets.
5-Point Cloud Processing – CloudCompare
The UAV LiDAR point clouds were post-processed in CloudCompare, an open-source software for large-scale geospatial datasets. The workflow included noise filtering, ground segmentation, height-normalisation, and the generation of digital terrain models (DTM) and digital surface models (DSM). CloudCompare’s powerful subsampling and scalar-field tools were used to prepare the landscape point clouds for integration into the combined 3D environment and for subsequent GIS-based morphological analysis.
8-Analysis and Visualisation – GIS
The integrated 3D datasets were imported into a GIS environment to support landscape-level analysis and spatial interpretation. Digital elevation models derived from LiDAR were analysed for slope, visibility, hydrology, and landscape connectivity. This allowed the surroundings of the tower to be studied in relation to its topographic position, historical routes, and broader environmental context. GIS served as the analytical bridge between high-resolution 3D documentation and archaeological interpretation.
3-Photogrammetric Recording – Image-Based Modelling
A complementary image-based modelling campaign was carried out using high-resolution RGB photographs acquired from both ground level and UAV platforms. These images were used to capture detailed colour textures and to support the reconstruction of areas with complex reflectance or occlusions. The photogrammetry dataset was later incorporated into RealityScan to enhance colour fidelity and improve texturing across the 3D models. This combination of laser-scanned geometry and photographic data ensured a visually accurate and metrically reliable representation of the monument.
6-Laser Scan Refinement – MeshLab
The laser-scanning datasets underwent an initial refinement phase in MeshLab, where passes were trimmed, small artefacts were removed, and uniform meshing strategies were applied to maintain consistency across the model. MeshLab’s tools for smoothing, simplification, and hole-filling were used selectively to produce a clean but metrically consistent mesh suitable for merging with photogrammetric outputs and aerial data within RealityCapture.
9-Online Publication – 3DHOP
Once processed, the final 3D models were prepared for web dissemination using 3DHOP (https://3dhop.net/), a WebGL-based platform for high-resolution 3D visualisation. The models were converted into multiresolution NEXUS format, enabling smooth online navigation, real-time lighting controls, orbit/pan/zoom, measurement tools, and sectioning functions. This made the datasets accessible to researchers without specialised 3D software, supporting remote collaboration, teaching, and long-term digital preservation.