Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DeepAerialMapper: Deep Learning-based Semi-automatic HD Map Creation for Highly Automated Vehicles

Robert Krajewski, Huijo Kim

TL;DR

HD maps are essential for highly automated vehicles but costly to produce from ground-based data. This paper presents DeepAerialMapper, a semi-automatic workflow that semantically segments high-resolution aerial imagery and post-processes contours to generate Lanelet2 HD-map prototypes. It introduces a new Aachen dataset with eight classes, demonstrates a CNN-based segmentation plus a symbol classifier, and a contour-based map extraction achieving 96-97% recall/precision relative to manual references. The Lanelet2 export enables easy extension with standard tools, suggesting aerial imagery can substantially reduce manual map creation effort for HD maps.

Abstract

High-definition maps (HD maps) play a crucial role in the development, safety validation, and operation of highly automated vehicles. Efficiently collecting up-to-date sensor data from road segments and obtaining accurate maps from these are key challenges in HD map creation. Commonly used methods, such as dedicated measurement vehicles and crowd-sourced data from series vehicles, often face limitations in commercial viability. Although high-resolution aerial imagery offers a cost-effective or even free alternative, it requires significant manual effort and time to transform it into maps. In this paper, we introduce a semi-automatic method for creating HD maps from high-resolution aerial imagery. Our method involves training neural networks to semantically segment aerial images into classes relevant to HD maps. The resulting segmentation is then hierarchically post-processed to generate a prototypical HD map of visible road elements. Exporting the map to the Lanelet2 format allows easy extension for different use cases using standard tools. To train and evaluate our method, we created a dataset using public aerial imagery of urban road segments in Germany. In our evaluation, we achieved an automatic mapping of lane markings and road borders with a recall and precision exceeding 96%. The source code for our method is publicly available at https://github.com/RobertKrajewski/DeepAerialMapper.

DeepAerialMapper: Deep Learning-based Semi-automatic HD Map Creation for Highly Automated Vehicles

TL;DR

HD maps are essential for highly automated vehicles but costly to produce from ground-based data. This paper presents DeepAerialMapper, a semi-automatic workflow that semantically segments high-resolution aerial imagery and post-processes contours to generate Lanelet2 HD-map prototypes. It introduces a new Aachen dataset with eight classes, demonstrates a CNN-based segmentation plus a symbol classifier, and a contour-based map extraction achieving 96-97% recall/precision relative to manual references. The Lanelet2 export enables easy extension with standard tools, suggesting aerial imagery can substantially reduce manual map creation effort for HD maps.

Abstract

High-definition maps (HD maps) play a crucial role in the development, safety validation, and operation of highly automated vehicles. Efficiently collecting up-to-date sensor data from road segments and obtaining accurate maps from these are key challenges in HD map creation. Commonly used methods, such as dedicated measurement vehicles and crowd-sourced data from series vehicles, often face limitations in commercial viability. Although high-resolution aerial imagery offers a cost-effective or even free alternative, it requires significant manual effort and time to transform it into maps. In this paper, we introduce a semi-automatic method for creating HD maps from high-resolution aerial imagery. Our method involves training neural networks to semantically segment aerial images into classes relevant to HD maps. The resulting segmentation is then hierarchically post-processed to generate a prototypical HD map of visible road elements. Exporting the map to the Lanelet2 format allows easy extension for different use cases using standard tools. To train and evaluate our method, we created a dataset using public aerial imagery of urban road segments in Germany. In our evaluation, we achieved an automatic mapping of lane markings and road borders with a recall and precision exceeding 96%. The source code for our method is publicly available at https://github.com/RobertKrajewski/DeepAerialMapper.
Paper Structure (17 sections, 4 figures, 3 tables)

This paper contains 17 sections, 4 figures, 3 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Data Acquisition
  4. Processing Aerial Imagery
  5. Lanelet2 Format
  6. SEMANTIC SEGMENTATION OF AERIAL IMAGERY
  7. SYMBOL CLASSIFICATION
  8. CREATION OF HD MAPS
  9. Road Border
  10. Lane Markings
  11. Lanes
  12. Exporting as Lanelet2 Map
  13. EXPERIMENTS
  14. Semantic Segmentation
  15. Symbol Classification
  16. ...and 2 more sections

Figures (4)

  • Figure 1: Exemplary semantic segmentation of an intersection shown as overlay. The annotated classes are represented by colours (Red: symbols, cyan: lane markings, gray: road, green: vegetation, blue: walkway, purple: traffic island)
  • Figure 2: Exemplary result of road border splitting. Different colours represent individual road borders.
  • Figure 3: Exemplary result of symbol classification.
  • Figure 4: Mimicking occlusion of symbols by cropping and cutting out random parts.