BEV-SLD: Self-Supervised Scene Landmark Detection for Global Localization with LiDAR Bird's-Eye View Images
🤖 AI Summary
This work addresses the lack of scene-specific semantic cues in LiDAR-based global localization by proposing a novel self-supervised approach that introduces self-supervised scene landmark detection to this task for the first time. Leveraging LiDAR bird’s-eye-view representations, the method automatically discovers spatially consistent, scene-specific patterns as semantic landmarks without requiring manual annotations. A consistency loss aligns learnable global landmark coordinates with per-frame heatmaps to enable robust localization. Experimental results demonstrate that the proposed method achieves superior and robust global localization performance across diverse environments—including campus, industrial, and forest settings—significantly outperforming current state-of-the-art approaches.
Application Category
📝 Abstract
We present BEV-SLD, a LiDAR global localization method building on the Scene Landmark Detection (SLD) concept. Unlike scene-agnostic pipelines, our self-supervised approach leverages bird's-eye-view (BEV) images to discover scene-specific patterns at a prescribed spatial density and treat them as landmarks. A consistency loss aligns learnable global landmark coordinates with per-frame heatmaps, yielding consistent landmark detections across the scene. Across campus, industrial, and forest environments, BEV-SLD delivers robust localization and achieves strong performance compared to state-of-the-art methods.
Problem
Research questions and friction points this paper is trying to address.
global localization
LiDAR
scene landmark detection
bird's-eye view
self-supervised
Innovation
Methods, ideas, or system contributions that make the work stand out.
Self-supervised learning
Scene Landmark Detection
Bird's-Eye View
LiDAR localization
Consistency loss
🔎 Similar Papers
No similar papers found.
