Robust Self-Tuning Data Association for Geo-Referencing Using Lane Markings

TL;DR

The paper addresses robust geo-referencing localization by leveraging lane markings as dual-view landmarks and introducing a self-tuning data association that adapts the search space based on a pseudo-entropy measure of the lane-marking information. It couples this front-end adaptation with dynamic covariance adjustment, propagating pose uncertainty from data association to detections to smooth the trajectory. The core contributions are a complete pipeline for rural and urban geo-referencing, the DA-LMR–based pseudo-entropy metric, the DC-SAC–driven self-tuning search, and a covariance propagation scheme tied to the relative transform results. Experimental results in Karlsruhe show improved outlier mitigation and smoother trajectories compared to state-of-the-art methods, with particular gains in high-aliasing, straight-road regimes relevant to rural environments.

Abstract

Localization in aerial imagery-based maps offers many advantages, such as global consistency, geo-referenced maps, and the availability of publicly accessible data. However, the landmarks that can be observed from both aerial imagery and on-board sensors is limited. This leads to ambiguities or aliasing during the data association. Building upon a highly informative representation (that allows efficient data association), this paper presents a complete pipeline for resolving these ambiguities. Its core is a robust self-tuning data association that adapts the search area depending on the entropy of the measurements. Additionally, to smooth the final result, we adjust the information matrix for the associated data as a function of the relative transform produced by the data association process. We evaluate our method on real data from urban and rural scenarios around the city of Karlsruhe in Germany. We compare state-of-the-art outlier mitigation methods with our self-tuning approach, demonstrating a considerable improvement, especially for outer-urban scenarios.

Figures (8)

• Figure 1: The complete pipeline of our self-tuning geo-referencing approach.
• Figure 2: Example of 2D projected polylines for a set of detections $\mathcal{D}_i$. We mark in colors four polylines that we represent as 1D signals in Fig. \ref{['fig:delta']}.
• Figure 3: Example of 1D signals $\mathbf{\Delta}_h$ derived from polylines represented in Fig. \ref{['fig:polylines']} with different colors.
• Figure 4: Covariance propagation in a set $\mathcal{D}_i$, where the red ellipse indicates the data association covariance $\Sigma_{i}$ and the green ellipses depict the detections covariances propagated $\Sigma_{i_k}$.
• Figure 5: The four closed trajectories for evaluation, driving through the city of Karlsruhe (Germany) and its outer roads. In Table \ref{['tab:datasets']}, we show more details about these trajectories.