Feasibility and Accuracy of as-Built Modelling From Slam-Based Point Clouds: Preliminary Results

Abstract

Abstract. Nowadays, portable Mobile Mapping Systems (MMSs) and robotic mapping platforms leveraging on Simultaneous Localization and Mapping (SLAM) methods are gaining increasing attention for architectural and construction surveying, representing an efficient solution for geometric data acquisition for scan-to-BIM purposes. However, the applicability of standard modelling workflows and the accuracy of Building Information Models (BIM) that can be obtained from SLAM-based point clouds is still an open question. In this paper, we propose a preliminary evaluation on the feasibility of extracting as-built BIM from (i) a point cloud acquired with a commercial portable MMS, and (ii) a point cloud obtained through an open-source SLAM algorithm, surveying the environment with an autonomous mobile robotic platform. In both cases, the main structural elements of the test site are accurately generated, thus showing promising results. On the other hand, the experiment highlights also the need for SLAM systems capable of providing less noisy point clouds, in order to capture and model architectural details.

BIMBot for Autonomous Laser Scanning in Built Environments

“…While "scanning-whilenavigating" approaches like those in [19,20] are efficient, they rely heavily on SLAM algorithms, which produce noisy point clouds. This makes them unsuitable for architectural detail capture, unlike automated TLS-based approaches [12][13][14].…”

“…Other studies [12][13][14][15] have explored different scanning approaches and devices for evaluating point cloud quality in 3D as-built BIM model creation. These studies have suggested that point clouds generated by low-quality LiDAR devices using SLAM are often too sparse and inaccurate.…”