Assessing the Value of 3D Reconstruction in Building Construction

3-dimensional (3D) reconstruction is an emerging field in image processing and computer vision that aims to create 3D visualizations/ models of objects/ scenes from image sets. However, its commercial applications and benefits are yet to be fully explored. In this paper, we describe ongoing work towards assessing the value of 3D reconstruction in the building construction domain. We present preliminary results from a user study, where our objective is to understand the use of visual information in building construction in order to determine problems with the use of visual information and identify potential benefits and scenarios for the use of 3D reconstruction.

💡 Research Summary

The paper presents early results from an ongoing investigation into the practical value of image‑based three‑dimensional (3D) reconstruction for building construction. While photographs, drawings, BIM models, and video are routinely used to convey site information, the authors argue that current visual information suffers from three systemic shortcomings: (1) miscommunication due to differing interpretations among owners, architects, contractors, and subcontractors; (2) incompleteness or inaccuracy because legacy design documents and photos often do not reflect as‑built conditions; and (3) a lack of spatial orientation, making it difficult for stakeholders to understand the geometry of a site without repeated physical visits.

To uncover these issues, the researchers conducted semi‑structured interviews with two construction professionals—a renovation estimator (P1) and a project manager (P2). Each interview lasted about an hour and was analyzed using open coding. The participants confirmed that visual misinterpretation leads to rework, cost overruns, and schedule delays. Specific anecdotes include a historic building renovation where 30,000 original drawings would have been required for a new build, and a skylight installation where design dimensions differed from the actual concrete opening, forcing costly replacements.

The authors propose that 3D reconstruction from ordinary photographs can mitigate these problems. By processing a set of images into a dense point‑cloud model, users obtain an interactive, geo‑referenced 3D representation of the site. This model enables (a) immersive spatial awareness—users can “walk” through the scene, view it from arbitrary viewpoints, and zoom to inspect fine details; (b) accurate measurement—real‑world dimensions can be extracted directly, supporting prefabrication of components such as custom skylights; and (c) reduced need for on‑site visits, because the model supplies up‑to‑date context for planning, coordination, and verification tasks.

The paper outlines several concrete scenarios where these capabilities would be beneficial: (i) renovation of acoustic ceilings, where a 3D model could clarify hidden utilities and prevent measurement errors; (ii) site orientation for new foundations, allowing engineers to assess soil levels and stone strata virtually; and (iii) prefabrication of custom parts, where precise as‑built dimensions reduce material waste and labor.

Importantly, the authors position image‑based reconstruction as a potentially lower‑cost alternative to laser scanning (LiDAR). While LiDAR offers high precision, it requires specialized equipment and expertise; photogrammetry can be performed with standard cameras and off‑the‑shelf software, making it more accessible for many construction firms.

Future work will move from qualitative interviews to quantitative experiments. The researchers plan to design tasks that compare the performance of participants using traditional visual artifacts (2D photos, drawings, BIM) against those using 3D reconstructions. They intend to isolate immersion components—field‑of‑view, field‑of‑regard, and zoom level—and measure their impact on task speed, error rate, and perceived workload. This approach mirrors prior studies in tele‑operation that demonstrated superior performance for 3D visualizations over video feeds.

In summary, the paper identifies critical gaps in current visual communication on construction sites, proposes photogrammetric 3D reconstruction as a versatile solution, and outlines a roadmap for rigorous evaluation. By linking identified problems to concrete benefits and experimental plans, the work lays a solid foundation for assessing whether 3D reconstruction can become a mainstream tool for improving efficiency, accuracy, and collaboration in the building construction industry.