USAMV BUCHAREST · STUDENT SYMPOSIUM 2026

Development of a Digital Twin
for the USAMV Bucharest Campus

Using Terrestrial Scanning and Geospatial Technologies

A Comparative Analysis between Romania and Nordic Countries

Authors: Roxana Maria Haican · Alexandra-Ionela Nistor · Nicoleta Pavel · Ana Maria Savu · Iulian-Sebastian Hagiu · Vlad-Andrei Haican Coord: Lect. PhD. Eng. Alexandru Calin Email: alenistor2002@gmail.com

Context · ObjectivesWhat is a Digital Twin?

A Digital Twin is a three-dimensional digital replica of a real-world environment, increasingly used in urban planning, infrastructure management, and academic research. While Nordic countries have systematically integrated such frameworks, Romania exhibits limited adoption — primarily due to financial constraints and restricted access to advanced geospatial infrastructure.

RESEARCH GOAL

Develop a detailed digital twin of the USAMV Bucharest campus

using cost-effective geospatial technologies, while comparing the methodology with Nordic practices.

Key numbersMission at a glance

UAV photos
419 geo-tagged
Flight duration
32.4 min 4.99 km path
SLAM poses
17,272 @ 9.1 Hz
Scan duration
31.5 min 1.66 km path
LiDAR points
6.5 M 14 PCD files
RMS error
< 3 cm after ICP
LOD
200–300 BIM-compatible
Cost reduction
60–80% vs. high-end TLS

State of the artNordic Reference Frameworks

National initiatives systematically integrate UAV photogrammetry and SLAM LiDAR into urban planning, achieving fit-for-purpose accuracy of ≈ 2–5 cm.

FINLAND

Helsinki 3D+

City-scale digital twin · UAV photogrammetry + mobile mapping + SLAM handheld scanners.

Agisoft Metashape · CloudCompare · GIS

DENMARK

GeoDanmark

Distributed geospatial acquisition · GNSS RTK · DJI UAV · high-res RGB sensors.

Pix4Dmapper · Metashape · ArcGIS Pro

SWEDEN

Lantmäteriet

Mobile LiDAR + UAV (Phantom, Matrice) · GNSS/IMU continuous capture along corridors.

SfM in Metashape · CloudCompare

NORWAY

Kartverket

National mapping authority · UAV + terrestrial LiDAR · open-data culture.

Standardized 2–5 cm urban accuracy

WorkflowIntegrated Multi-Sensor Methodology

01

Acquisition

Aerial UAV + Terrestrial SLAM

02

Processing

SfM/MVS in Metashape · SLAM optimization in FJD Trion Studio

03

Alignment

Coarse + ICP fine registration in CloudCompare

04

Digital Twin

Unified 3D model · LOD 200–300 · RMS < 3 cm

UAV PhotogrammetryAerial Acquisition — DJI Mini 2

Equipment specifications

Weight< 249 g
Camera12 MP · 4000 × 3000 px
Focal length4.49 mm (≈ 24 mm equiv.)
Aperture / ISOf/2.8 · ISO 100
Stabilization3-axis mechanical gimbal
PositioningGNSS (GPS + GLONASS), no RTK/PPK
Geo-referencingvia Ground Control Points (GCPs)
Flight strategyNadir 30–50 m + oblique 45°
Image overlap70–80% forward & lateral
UAV flight track over USAMV
Fig. 1 — Aerial flight track over USAMV campus (Esri imagery)
Altitude profile
Fig. 2 — Altitude profile (stable plateau ≈ 206 m MSL ≈ 135 m AGL)

SLAM LiDARTerrestrial Acquisition — FJD Trion S2

Equipment specifications

Weight2 kg (with battery)
LiDAR sensormulti-line laser
Field of view360° × 270°
Rangeup to 120 m (80% reflectivity)
Capture rate640.000 pts / s
Relative accuracyup to 1.2 cm
Cameras2 × 12 MP panoramic
PositioningSLAM + IMU + VIO + RTK / PPK
Operating modehandheld, walking 0.5–1.0 m/s
SLAM trajectory top view
Fig. 3 — SLAM trajectory · 17,272 poses · color = time (start → end)
Point cloud top view
Fig. 4 — LiDAR point cloud (top) overlaid with white scanner trajectory
Point cloud 3D
Fig. 5 — LiDAR point cloud — isometric 3D view

Visualization ModesPoint Cloud — RGB vs Intensity

Two complementary visualization modes derived from the unified point cloud — RGB (left of each pair) for visual realism, intensity (right) for material analysis.

Top-view RGB vs intensity
Fig. 6 — Top-view aerial · RGB ← → intensity
Aerial façade RGB vs intensity
Fig. 7 — Façade aerial · RGB ← → intensity
Façade close-up RGB vs intensity
Fig. 8 — Façade close-up · RGB ← → intensity
VISUALIZATION INSIGHT

RGB visualization preserves true color from camera images — useful for visual interpretation and presentations. Intensity visualization reflects LiDAR return strength — highlights material reflectivity and structural details invisible in RGB.

Acquisition SoftwareFJD Trion Studio — Real-time Interface

During the campus scan, the FJD Trion Studio app on the device shows a live floor-plan view of the SLAM trajectory together with the dual fisheye camera feeds.

FJD Trion Studio interface
Fig. 9 — FJD Trion Studio: floor-plan view (top) and dual camera feeds (bottom)
FJD Trion S2 in the field
Fig. 10 — FJD Trion S2 deployed on USAMV grounds

Interface features

  • Real-time floor-plan view
  • Live trajectory tracking (green path)
  • Dual camera preview (Left + Right)
  • On-site quality validation
  • Loop closure indication
Video — Workflow demonstration (from acquisition through processing)

Live viewer · 2D mapInteractive Orthophoto

Geo-referenced orthophoto of the USAMV / FIFIM area, served on the UGA Aerial Point Cloud Hub.

Open in new tab → 2D map view · pan/zoom · measurements · Stereo 70 (EPSG:3844) · mobile-friendly

Live viewer · 3D point cloudInteractive Point Cloud — Potree

Full LiDAR point cloud of the FIFIM Faculty viewable in any web browser. Includes measurement, sections, and georeferenced PNG export.

Open in new tab → 3D rotate / zoom · measurements · sections · PNG+PGW export · Solid Surface mode

Economic ImpactCost & Time Efficiency

Equipment cost reduction

60 – 80 %

vs. high-end terrestrial laser scanners (Leica · FARO · Riegl)

Acquisition time reduction

50 – 65 %

vs. conventional total-station + static TLS workflow

Key takeawaysConclusions

01

FEASIBILITY

An accurate, scalable digital twin can be built in Romania with consumer-grade equipment.

02

COMPLEMENTARITY

Multi-sensor fusion (UAV + SLAM LiDAR) overcomes occlusion and coverage limits.

03

ADOPTABILITY

Methodology aligned with Nordic practices — directly transferable to Romanian institutional contexts.

Mobile · Live web viewersOpen the live data on your phone

Scan a QR code below to open the orthophoto map or the LiDAR point cloud directly in your phone's browser.

Orthophoto map

WebODM result · Stereo 70 (EPSG:3844)

QR Orthophoto

drone.ugaerial.ro/Orto_-_Facultate_Fifim_-_Test.tif/map

LiDAR point cloud

Potree 3D viewer · UAV scan

QR LiDAR

drone.ugaerial.ro/LAZ_LIDAR_DRONA_FIFIM.las

BONUS · Our most recent experiment Our best approach — 3D Gaussian Splatting

Beyond the established LiDAR + photogrammetry pipeline, our team also explored a state-of-the-art neural rendering technique: 3D Gaussian Splatting (3DGS). From a small set of input photographs, 3DGS produces a photorealistic, fully navigable 3D scene in the browser — preserving real lighting, textures, and even reflections. The interactive scene below is hosted on superspl.at and can be explored directly in your browser.

Open in new tab → WebGL-powered · mobile-ready · zero plugins

Navigation · Mouse + Keyboard

Left drag Orbit / rotate the scene Right drag Pan (slide camera sideways) Scroll wheel Zoom in / out Double-click Focus on the clicked point W A S D Walk forward / left / back / right (FPS mode) Q  /  E Move down / up vertically Shift Move faster while pressed Space Reset / re-center camera

Side toolbar (inside the viewer)

⛶ Fullscreen — expand the viewer to the whole screen 🏠 Home view — reset to the original camera 🎬 Auto-rotate — let the camera orbit on its own 🥽 VR / AR — open in immersive mode (if device supports) ⚙ Quality — toggle render quality / performance ℹ Info — scene metadata, author, sharing options

On mobile: one-finger drag = orbit · two-finger pinch = zoom · two-finger drag = pan.

Audience feedback · AnonymousRate this presentation

No account, no email — just one click. Each device can vote once.

…
Views
…
Votes
—
Average / 10

How would you rate this presentation? Click a number from 1 (poor) to 10 (excellent).