Deep learning applications for point clouds in the construction industry

Abstract

Deep learning (DL) on point clouds holds significant potential in the construction industry, yet no comprehensive review has thoroughly summarized its applications and shortcomings. This paper presents a detailed review of the current applications of DL on point clouds in the construction industry, highlighting existing challenges, limitations, and future research directions. A two-stage literature search was conducted, resulting in the collection of 55 research papers published since 2020. The review provides an overview of DL algorithms and examines the datasets used for DL on point clouds, including both real-world and synthetic datasets. Furthermore, it summarizes the various applications of DL on point clouds within the construction sector. Following this analysis, the paper discusses current deficiencies and potential improvements in model performance and data-related issues. Finally, several recommendations are provided to advance the development of DL-based point cloud applications in the construction industry.