Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Analyzing the impact of semantic LoD3 building models on image-based vehicle localization

Antonia Bieringer, Olaf Wysocki, Sebastian Tuttas, Ludwig Hoegner, Christoph Holst

TL;DR

This work tackles GNSS-denied urban vehicle localization by leveraging image features aligned to high-fidelity semantic LoD3 building models. It develops a pipeline that generates virtual LoD model views, extracts and maps 2D features to 3D coordinates, and estimates vehicle trajectories via spatial resection, comparing LoD2 and LoD3 implementations. The results show LoD3 can increase feature availability and, in several scenarios, improve positioning accuracy, particularly where facades on both sides of streets provide rich cues; however, topology and underpass issues can limit gains. The study demonstrates a practical application of LoD3 models for map-based car positioning and suggests future work including SLAM integration and improved feature discrimination to reduce false matches in GNSS-denied urban canyons.

Abstract

Numerous navigation applications rely on data from global navigation satellite systems (GNSS), even though their accuracy is compromised in urban areas, posing a significant challenge, particularly for precise autonomous car localization. Extensive research has focused on enhancing localization accuracy by integrating various sensor types to address this issue. This paper introduces a novel approach for car localization, leveraging image features that correspond with highly detailed semantic 3D building models. The core concept involves augmenting positioning accuracy by incorporating prior geometric and semantic knowledge into calculations. The work assesses outcomes using Level of Detail 2 (LoD2) and Level of Detail 3 (LoD3) models, analyzing whether facade-enriched models yield superior accuracy. This comprehensive analysis encompasses diverse methods, including off-the-shelf feature matching and deep learning, facilitating thorough discussion. Our experiments corroborate that LoD3 enables detecting up to 69\% more features than using LoD2 models. We believe that this study will contribute to the research of enhancing positioning accuracy in GNSS-denied urban canyons. It also shows a practical application of under-explored LoD3 building models on map-based car positioning.

Analyzing the impact of semantic LoD3 building models on image-based vehicle localization

TL;DR

This work tackles GNSS-denied urban vehicle localization by leveraging image features aligned to high-fidelity semantic LoD3 building models. It develops a pipeline that generates virtual LoD model views, extracts and maps 2D features to 3D coordinates, and estimates vehicle trajectories via spatial resection, comparing LoD2 and LoD3 implementations. The results show LoD3 can increase feature availability and, in several scenarios, improve positioning accuracy, particularly where facades on both sides of streets provide rich cues; however, topology and underpass issues can limit gains. The study demonstrates a practical application of LoD3 models for map-based car positioning and suggests future work including SLAM integration and improved feature discrimination to reduce false matches in GNSS-denied urban canyons.

Abstract

Numerous navigation applications rely on data from global navigation satellite systems (GNSS), even though their accuracy is compromised in urban areas, posing a significant challenge, particularly for precise autonomous car localization. Extensive research has focused on enhancing localization accuracy by integrating various sensor types to address this issue. This paper introduces a novel approach for car localization, leveraging image features that correspond with highly detailed semantic 3D building models. The core concept involves augmenting positioning accuracy by incorporating prior geometric and semantic knowledge into calculations. The work assesses outcomes using Level of Detail 2 (LoD2) and Level of Detail 3 (LoD3) models, analyzing whether facade-enriched models yield superior accuracy. This comprehensive analysis encompasses diverse methods, including off-the-shelf feature matching and deep learning, facilitating thorough discussion. Our experiments corroborate that LoD3 enables detecting up to 69\% more features than using LoD2 models. We believe that this study will contribute to the research of enhancing positioning accuracy in GNSS-denied urban canyons. It also shows a practical application of under-explored LoD3 building models on map-based car positioning.
Paper Structure (19 sections, 7 equations, 7 figures, 6 tables)

This paper contains 19 sections, 7 equations, 7 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Semantic 3D city models
  4. Image-Based Vehicle Positioning
  5. Methodology
  6. Creating Virtual Images of the LoD Models
  7. Feature Matching
  8. Correlating the Features with a 3D-Coordinate
  9. Calculating the Trajectory
  10. Experiments
  11. Different Approaches for Feature Matching
  12. Corresponding Image-Pairs of Different Types
  13. Segmented Images with Deep Learning
  14. Sobel-Filter and Canny Edge Detector
  15. Results and Discussion
  16. ...and 4 more sections

Figures (7)

  • Figure 1: Developed workflow: The images of the real world are matched with virtual images of the LoD models. The combination of both will be used to calculate the trajectory.
  • Figure 2: Virtual setup for ray casting.
  • Figure 3: Ray casting results for LoD3 a) hit distance b) geometry IDs c) primitive IDs d) primitive normals e) barycentric coordinates.
  • Figure 4: Workflow of the generation of the feature images.
  • Figure 5: Visualization of the test area divided into three sub-areas (purple) with different availability of building models: a) both in LoD2 and LoD3 (blue), and b) solely in LoD2 (green).
  • ...and 2 more figures