Data for 3D reconstruction and point cloud classification using machine learning in cultural heritage environment

Abstract

Unmanned Aerial Vehicle (UAV) photogrammetry, thanks to the development of Structure from Motion (SfM) and Multi-View Stereo (MVS) algorithms, allows the generation of dense point clouds, capable of representing three-dimensional objects and structures in a detailed and accurate manner. In addition, the possibility of associating more semantic information through automatic segmentation and classification models, becomes of fundamental importance in the field of development, protection and maintenance of Cultural Heritage (CH). With the developments in Artificial Intelligence (AI), classification algorithms based on Machine Learning (ML) have been developed. In particular, the Random Forest is used in order to perform a semantic classification of the point cloud generated by UAV photogrammetry and Global Navigation Satellite Systems (GNSS) survey of a structure belonging to CH environment. Indeed, this paper describes the images collected through a UAV survey, for 3D reconstruction of Temple of Hera (Italy) based on photogrammetric approach and georeferenced by the use of 8 Ground Control Points (GCPs) acquired by GNSS survey. In addition, the shared dataset contains the point cloud and data for classification using Random Forest algorithm.