3D Mapping of Vacant Buildings for Reuse
PhD project by Povl Filip Sonne-Frederiksen
The project titled “3D Mapping of Vacant Buildings for Reuse” seeks to tackle a critical challenge in the architecture, construction and engineering (AEC) industry: efficiently repurposing existing structures. In an era where sustainable development and environmental responsibility are paramount, the ability to reuse vacant buildings instead of demolishing them can significantly contribute to a circular economy. However, one of the primary barriers to reusing old buildings is the lack of accurate documentation and blueprints, which are often missing, outdated, or nonexistent. The research aims to overcome these hurdles by developing an innovative technology pipeline that allows for quick and user-friendly mapping of buildings, providing essential data for early-phase building reuse analysis.
Background and Significance
Reusing existing buildings offers substantial benefits, including reduced waste, decreased demand for new construction materials, and preservation of historical structures. In some countries, such as Norway, there is a legal requirement to conduct an Ombrukskartlegging (reuse mapping) for any building slated for demolition or transformation, ensuring that reusable components are identified and salvaged. Traditionally, this process is manual, time-consuming, and prone to errors. This project introduces a novel approach that leverages technologies like LiDAR and machine learning to construct a 3D modelling to automate and streamline this process, making it more efficient and reliable.
Research Objectives
The primary objectives of this research are:
1. Develop a Scan-to-3D Pipeline: Create a seamless process that captures detailed 3D models of buildings using affordable devices like iPads or iPhones with LiDAR sensors.
2. Facilitate Early-Phase Reuse Analysis: Provide stakeholders with essential data that can inform decisions about a building’s potential reuse during the early stages of design.
3. Integrate with Existing Tools: Ensure compatibility with existing environmental analysis tools to evaluate environmental impact and potential energy efficiency improvements.
Methodology
The core of this research lies in the development of a Scan-to-BIM (Building Information Modeling) pipeline. This process begins with capturing the physical structure of a building using a LiDAR sensor and camera on an iPad or iPhone. The captured data is transformed into a 3D “point cloud,” a digital representation of the building’s spatial dimensions, which is further processed into a simplified building model and point clouds of labelled loose inventory. Here’s a closer look at the steps involved:
- Data Collection: A LiDAR sensor and camera are used to scan the building using relatively low-cost devices like an iPad or iPhone Pro. The device’s position and orientation data are combined with the captured images to form a 3D point cloud.
- Object Segmentation and Labeling: The captured images are processed using a machine-learning algorithm that identifies and labels objects within the building. This step is crucial for recognizing loose inventory.
- 3D Model Construction: The point cloud data is converted into a simplified 3D model using planar region growing. This method involves analyzing the point cloud to find flat surfaces, constructing a 3D cell complex, and using a linear integer solver to create a coherent selection of surfaces, resulting in a clean model.
- Integration with Analytical Tools: The resulting 3D model is exported to platforms like Speckle, enabling further analysis and integration with workflows like Speckle Automate or Grasshopper. This integration allows for environmental impact assessments and other evaluations essential for reuse planning.
Case Studies and Findings
The effectiveness of this pipeline is demonstrated through two critical case studies:
- Tvinn Solutions Integration: This case study focuses on integrating the pipeline with Tvinn Solutions, which generates digital twins of building components in Revit. The goal is to evaluate the accuracy and completeness of the labelled geometry produced by the machine learning algorithm and identify any additional metadata require.
- Environmental Impact Assessment: Using tools like Ladybug Tools, Radiance, and SpeckleLCA, this case study assesses whether the simplified geometry generated by the pipeline provides enough detail for meaningful environmental analysis. The results show that the simplified models are suitable for early-phase assessments, enabling stakeholders to make informed decisions quickly.
KONTAKT
Relateret
MÅSKE ER DU OGSÅ INTERESSERET I DETTE
RELATERET INDHOLD
