Related Experiment Video
Updated: Sep 22, 2025
06:54
Photorealistic Learned Landscapes for Augmented Reality
Published on: June 27, 2025
211
Data for 3D reconstruction and point cloud classification using machine learning in cultural heritage environment
Massimiliano Pepe1, Vincenzo Saverio Alfio1, Domenica Costantino1
1Polytechnic of Bari, via E. Orabona 4, Bari 70125, Italy.
Data in Brief
|May 23, 2022
Summary
Unmanned Aerial Vehicle (UAV) photogrammetry combined with machine learning classifies 3D cultural heritage structures. This approach enhances the protection and maintenance of historical sites like the Temple of Hera.
Area of Science:
- Geomatics and Geospatial Technology
- Artificial Intelligence in Cultural Heritage
- 3D Reconstruction and Digital Preservation
Background:
- Unmanned Aerial Vehicle (UAV) photogrammetry, utilizing Structure from Motion (SfM) and Multi-View Stereo (MVS), generates detailed 3D point clouds.
- Integrating semantic information via automatic segmentation and classification is crucial for Cultural Heritage (CH) development, protection, and maintenance.
- Advancements in Artificial Intelligence (AI) and Machine Learning (ML) enable sophisticated classification of 3D data.
Purpose of the Study:
- To semantically classify a 3D point cloud of the Temple of Hera using UAV photogrammetry and Global Navigation Satellite Systems (GNSS) data.
- To demonstrate the application of the Random Forest algorithm for detailed analysis of cultural heritage structures.
- To provide a shared dataset including point clouds and classification data for further research.
Main Methods:
- Acquisition of images via UAV survey for 3D reconstruction of the Temple of Hera.
- Georeferencing the 3D model using 8 Ground Control Points (GCPs) obtained through GNSS survey.
- Semantic classification of the generated point cloud using the Random Forest ML algorithm.
Main Results:
- Successful 3D reconstruction of the Temple of Hera using photogrammetric techniques.
- Generation of a dense, accurate point cloud representing the CH structure.
- Demonstration of Random Forest's efficacy in semantically classifying the UAV-derived point cloud.
Conclusions:
- UAV photogrammetry coupled with ML offers a powerful tool for detailed 3D analysis and semantic classification of cultural heritage sites.
- The developed methodology and dataset facilitate enhanced understanding, protection, and maintenance strategies for CH.
- This research contributes to the digital preservation of cultural heritage through advanced geospatial and AI techniques.
