Paper ThAT5.2
Eisemann, Leon (Porsche Engineering Group GmbH), Maucher, Johannes (Stuttgart Media University)
A NeRF-based Approach for Monocular Traffic Sign Reconstruction and Localization
Scheduled for presentation during the Regular Session "Sensing, Vision, and Perception III" (ThAT5), Thursday, September 26, 2024,
10:50−11:10, Salon 13
2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada
This information is tentative and subject to change. Compiled on October 8, 2024
|
|
Keywords Sensing, Vision, and Perception, Simulation and Modeling, Accurate Global Positioning
Abstract
High-resolution maps are a crucial factor for the functionality of highly automated driving functions. These contain precise information about lane geometry, road conditions, and traffic signs. Although the importance of high-definition maps, creating these is still an ongoing research topic. While the extraction of road geometry and lane information has seen significant progress, the reconstruction of traffic signs is in its early stages. We propose a novel approach for the accurate 3D localization and reconstruction of directional traffic signs, based on monocular images. For a flexible and extensible process, we use a neural radiance field based reconstruction pipeline. We demonstrate that our approach can localize signs with high precision, outperforming current methods, while also generating 3D representations, for use in mapping or simulation applications.
|
|