Cross-View Yaw Estimation in Location Uncertainty with Line-Aligning Yaw Scoring

2026-06-20

AI Summary

This work addresses the challenge of inaccurate yaw estimation in cross-view localization caused by positional uncertainty and large yaw ambiguities. The proposed method, LAYS, achieves sub-degree yaw estimation without requiring precise positional priors—a first in the field. By decoupling yaw estimation from positional accuracy, LAYS introduces a radially invariant line-consistency voting mechanism that aggregates feature matches between ground-view image columns and bird’s-eye-view (BEV) pixels within a discretized 3D pose space. Robust yaw estimates are then obtained via peak detection in the voting space. Evaluated on Mapillary, Ford, KITTI, and VIGOR datasets under standard field-of-view conditions with unknown yaw, LAYS improves yaw estimation accuracy by 28–45 percentage points, substantially enhancing downstream 3-DoF localization performance.

Abstract

Accurate yaw estimation is a bottleneck in cross-view localization between ground view and Bird's Eye View (BEV). Existing methods couple yaw with translation and rely on height or projection assumptions that degrade under large yaw ambiguity. We disentangle yaw from location accuracy and introduce LAYS, a radially invariant line-consensus voting method. By exploiting the radial invariance of our formulation, we achieve sub-degree yaw precision via 3D voting over all candidate poses, while eliminating the need for accurate location. Our key observation is that a ground-image column matched to BEV pixels induces the same yaw across all camera positions along the radial direction of the pixels. LAYS matches BEV pixels to ground columns using feature similarity and accumulates the induced yaw votes into discrete 3D bins, where correct correspondences along the radial line concentrate into a sharp peak for the correct yaw. Experiments on Mapillary, Ford, KITTI, and VIGOR show significant gains under unknown yaw, particularly for normal FoV with unknown yaw (+28$\sim$45\%p), and using LAYS as a yaw prior improves downstream 3-DoF localization.

Problem

• cross-view localization
• yaw estimation
• location uncertainty
• Bird's Eye View
• radial invariance

Innovation

• yaw estimation
• cross-view localization
• radial invariance
• 3D voting
• line-consensus