B$^2$F-Map: Crowd-sourced Mapping with Bayesian B-spline Fusion
🤖 AI Summary
This work addresses the challenges of high-definition (HD) map construction in crowdsourced settings, where prior maps are unavailable and uncertainty from heterogeneous trajectory sources is difficult to integrate. The authors propose a lane-level HD map generation method that relies solely on monocular cameras, consumer-grade GNSS, and IMUs commonly available in mass-produced vehicles. Locally, extended object tracking with Poisson multi-Bernoulli (EOT-PMB) filtering combined with Gibbs sampling is employed for map building. In the cloud, lane geometry is represented using B-splines parameterized by Gaussian control points, and a Bayesian B-spline fusion framework is designed to accommodate varying trajectory densities while preserving uncertainty awareness across multiple drive sessions. Extensive validation on large-scale real-world datasets demonstrates that the approach efficiently produces geometrically consistent HD maps suitable for diverse driving scenarios.
Application Category
📝 Abstract
Crowd-sourced mapping offers a scalable alternative to creating maps using traditional survey vehicles. Yet, existing methods either rely on prior high-definition (HD) maps or neglect uncertainties in the map fusion. In this work, we present a complete pipeline for HD map generation using production vehicles equipped only with a monocular camera, consumer-grade GNSS, and IMU. Our approach includes on-cloud localization using lightweight standard-definition maps, on-vehicle mapping via an extended object trajectory (EOT) Poisson multi-Bernoulli (PMB) filter with Gibbs sampling, and on-cloud multi-drive optimization and Bayesian map fusion. We represent the lane lines using B-splines, where each B-spline is parameterized by a sequence of Gaussian distributed control points, and propose a novel Bayesian fusion framework for B-spline trajectories with differing density representation, enabling principled handling of uncertainties. We evaluate our proposed approach, B$^2$F-Map, on large-scale real-world datasets collected across diverse driving conditions and demonstrate that our method is able to produce geometrically consistent lane-level maps.
Problem
Research questions and friction points this paper is trying to address.
crowd-sourced mapping
HD map generation
map fusion
uncertainty handling
lane-level mapping
Innovation
Methods, ideas, or system contributions that make the work stand out.
Bayesian B-spline fusion
crowd-sourced HD mapping
uncertainty-aware lane modeling
Poisson multi-Bernoulli filter
production vehicle sensing
🔎 Similar Papers
No similar papers found.
