University of Twente Student Theses


3D modelling of slums based on UAV data

Khawte, Sharvi Samir (2022) 3D modelling of slums based on UAV data.

[img] PDF
Abstract:Favelas are the most common type of informal settlements found in Brazil (Pedro & Queiroz, 2019). Considering the complexity of the built environment in favelas in São Paulo, mapping and assessing the information related to the morphology of the slums becomes a challenge. The Housing Secretariat, City Hall, São Paulo, has conducted surveys using Unmanned Aerial Vehicles (UAVs) for the favelas to visualize favelas in 3D, thus facilitating the slum upgrading projects. There is much research in developing a 3D city model for the urban environment, but limited research has been conducted on the 3D modelling of slums due to the irregular and unplanned houses. This study develops a methodological workflow for 3D building reconstructions in slums using Dense Image Matching (DIM) point clouds from UAVs and 2D cadastral data. This study also focuses on developing a workflow to create the 3D model and a Digital Twin (DT) of the slums by facilitating the data integration and updating the semantic information to provide additional information about individual buildings in the slums. Prior to the development of this research, semi-structured interviews were conducted to understand the stakeholders’ requirements and the challenges the experts are currently facing in working with the data for the slums. The point cloud was filtered into the ground and the non-ground (building) points. A noise filter algorithm was used to remove the noise or outliers in the point cloud, especially from the surface of the roofs. The automatic data-driven model approach for the 3D building reconstruction from point clouds and 2D cadastral data by Xiong et al. (2016) was selected as a baseline. The resulting 3D model was enriched with semantic information using an Extract, Transform, Load (ETL) platform. The result of the experiment showed irregular reconstructions in roof structures in some places in the 3D building models of the favela in the form of erroneous spikes. The errors found in the roof surface construction were due to the over-segmentation and the noise in the point cloud. A total of ten houses out of 2156 were not perfectly reconstructed as a result of occlusions due to the presence of vegetation. A qualitative evaluation of the developed workflow and the model was carried out using semi-structured interviews with the stakeholders to assess its feasibility in slum upgrading activities. The assessments concluded that the proposed workflow is suitable for creating digital twins for slums based on the UAV and 2D cadastral data. However, the 3D slum model had a few limitations, which are discussed in this research. During the semi-structured interviews to evaluate the model, four out of five interviewees agreed that the workflow could be used to visualize the buildings in slums in 3D and obtain information about individual buildings and update it in the future. However, due to the unavailability of ground truth or the validation data, the 3D model could not be evaluated quantitively. For further research, it is recommended to conduct field surveys to aid the comparison of the resulting 3D model with the validation/ground truth data. Keywords: Informal settlements/Slums, 3D building reconstruction/modelling, Digital Twins (DT), Unmanned Aerial Vehicles (UAV), Point Clouds
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page