Technological Trends in DL-Based 3D Digitization of Infrastructures by Using Point Clouds

Lizbeth C.; Ciro R. · Ingenierie des Systemes d'Information (2025) · DOI: 10.18280/isi.300323

Resumen

This article addresses the challenge of 3D digitization using point clouds, an essential process for technological trends such as Building Information Modeling (BIM) and digital twins. The process of capturing three-dimensional geometries from point clouds generated by laser scanning is a costly and time-consuming task. The purpose of the study is to analyze how deep learning (DL) models can optimize this process in different types of infrastructures (industrial, transportation, construction, and public and social services). To this end, a methodology based on a systematic review and a bibliometric analysis of the scientific literature indexed in Scopus was employed, examining advanced models such as PointNet, PointNet++, ResPointNet++, and datasets like CLOI, PSNet5, and Pipework. The results highlight the effectiveness of AI in the automated handling of point clouds, although limited documentation was identified in its application to mining infrastructures. In conclusion, a DL model and specific datasets for the efficient processing of point clouds in the mining sector are identified and recommended, thus contributing to the advancement of 3D digitization in this sector.

Instituciones: Universidad Nacional Mayor de San Marcos