Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LOD1 3D City Model from LiDAR: The Impact of Segmentation Accuracy on Quality of Urban 3D Modeling and Morphology Extraction

Fatemeh Chajaei, Hossein Bagheri

TL;DR

The paper demonstrates that LiDAR-derived DSMs, when combined with four deep semantic segmentation models and transfer learning, can produce accurate LOD1 3D building models and reliable urban morphology metrics using footprints extracted directly from LiDAR data. U-Net3+ emerges as the strongest footprint extractor after transfer learning, yielding improved 3D reconstructions and morphology estimates, particularly when height is computed with the median or $90^{th}$ percentile measures. The study shows a clear link between footprint segmentation accuracy (IoU) and the quality of height estimation, building-area, and exterior-wall calculations, highlighting the importance of robust footprint extraction for urban analysis. By avoiding multi-source data fusion and leveraging LiDAR alone, the approach offers a scalable, cost-effective solution for large-scale urban modeling and planning applications, with implications for energy, climate, and disaster-risk assessments.

Abstract

Three-dimensional reconstruction of buildings, particularly at Level of Detail 1 (LOD1), plays a crucial role in various applications such as urban planning, urban environmental studies, and designing optimized transportation networks. This study focuses on assessing the potential of LiDAR data for accurate 3D building reconstruction at LOD1 and extracting morphological features from these models. Four deep semantic segmentation models, U-Net, Attention U-Net, U-Net3+, and DeepLabV3+, were used, applying transfer learning to extract building footprints from LiDAR data. The results showed that U-Net3+ and Attention U-Net outperformed the others, achieving IoU scores of 0.833 and 0.814, respectively. Various statistical measures, including maximum, range, mode, median, and the 90th percentile, were used to estimate building heights, resulting in the generation of 3D models at LOD1. As the main contribution of the research, the impact of segmentation accuracy on the quality of 3D building modeling and the accuracy of morphological features like building area and external wall surface area was investigated. The results showed that the accuracy of building identification (segmentation performance) significantly affects the 3D model quality and the estimation of morphological features, depending on the height calculation method. Overall, the UNet3+ method, utilizing the 90th percentile and median measures, leads to accurate height estimation of buildings and the extraction of morphological features.

LOD1 3D City Model from LiDAR: The Impact of Segmentation Accuracy on Quality of Urban 3D Modeling and Morphology Extraction

TL;DR

The paper demonstrates that LiDAR-derived DSMs, when combined with four deep semantic segmentation models and transfer learning, can produce accurate LOD1 3D building models and reliable urban morphology metrics using footprints extracted directly from LiDAR data. U-Net3+ emerges as the strongest footprint extractor after transfer learning, yielding improved 3D reconstructions and morphology estimates, particularly when height is computed with the median or percentile measures. The study shows a clear link between footprint segmentation accuracy (IoU) and the quality of height estimation, building-area, and exterior-wall calculations, highlighting the importance of robust footprint extraction for urban analysis. By avoiding multi-source data fusion and leveraging LiDAR alone, the approach offers a scalable, cost-effective solution for large-scale urban modeling and planning applications, with implications for energy, climate, and disaster-risk assessments.

Abstract

Three-dimensional reconstruction of buildings, particularly at Level of Detail 1 (LOD1), plays a crucial role in various applications such as urban planning, urban environmental studies, and designing optimized transportation networks. This study focuses on assessing the potential of LiDAR data for accurate 3D building reconstruction at LOD1 and extracting morphological features from these models. Four deep semantic segmentation models, U-Net, Attention U-Net, U-Net3+, and DeepLabV3+, were used, applying transfer learning to extract building footprints from LiDAR data. The results showed that U-Net3+ and Attention U-Net outperformed the others, achieving IoU scores of 0.833 and 0.814, respectively. Various statistical measures, including maximum, range, mode, median, and the 90th percentile, were used to estimate building heights, resulting in the generation of 3D models at LOD1. As the main contribution of the research, the impact of segmentation accuracy on the quality of 3D building modeling and the accuracy of morphological features like building area and external wall surface area was investigated. The results showed that the accuracy of building identification (segmentation performance) significantly affects the 3D model quality and the estimation of morphological features, depending on the height calculation method. Overall, the UNet3+ method, utilizing the 90th percentile and median measures, leads to accurate height estimation of buildings and the extraction of morphological features.

Paper Structure

This paper contains 30 sections, 14 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Study Area and Datasets
  4. Study Area
  5. LiDAR Data
  6. OpenStreetMap Data
  7. Methods
  8. LiDAR Data Preparation
  9. Building Footprint Extraction
  10. Transfer Learning for Improving the Performance of Deep Segmentation Models
  11. Setting up Deep Segmentation Models
  12. Post-processing
  13. 3D Building Modeling at LOD1
  14. Building Morphology Extraction
  15. Evaluation Metrics
  16. ...and 15 more sections

Figures (14)

  • Figure 1: CityGML Levels of Detail (LOD0-4) representations RN130
  • Figure 2: A visualization of the study area, part of the Landsmeer, north of Amsterdam.
  • Figure 3: The framework used in this study for creating 3D models from LiDAR data and extracting urban morphological parameters.
  • Figure 4: Post-processing workflow for refining model outputs and generating building footprints
  • Figure 5: Workflow for generating 3D building models at LOD1 using LiDAR data and building footprints
  • ...and 9 more figures