POINT CLOUD CLASSIFICATION USING MACHINE LEARNING ALGORITHMS: APPLICATIONS IN CULTURAL HERITAGE ENVIRONMENT

3D point clouds are a source for generating 3D models of individual objects or cities due to the simplicity of acquiring and managing geospatial data. For this reason, 3D point cloud classification represents an interesting area of research due to its many advantages in different fields of application. Therefore, the aim of the paper is to identify a suitable methodology for semantic segmentation of the point cloud based on a multi-scale approach and machine learning algorithms. This methodology was applied on the following datasets belonging to the cultural heritage: Temple of Hera (Italy) and Sacro Monte di Ghiffa (Italy). The validation of the results obtained was carried out using appropriate performance indices (Precision, Overall Accuracy, Recall and F1-score) whose values showed the quality of the proposed method that is applicable to complex morphologies both at the scale of the individual artefact and at the urban scale. In addition, a comparison with approaches in the literature shows encouraging results in terms of both accuracy and speed of point cloud classification.