Sparse-Map LiDAR Place Recognition and Loop Closure

2025 - 2026
Zhejiang University

This project studies LiDAR place recognition and loop closure detection when the reference map is compact, incomplete, or intentionally sparse. I redesigned the sparse-map sampling strategy and evaluation protocol to better reflect deployment-oriented reference availability, including stricter keyframe selection, sparse-reference construction, and positive-reference analysis.

• Redesigned the sparse-reference sampling strategy for deployment-oriented LiDAR place recognition and loop closure evaluation.
• Refined keyframe selection, sparse-reference construction, and positive-reference analysis protocols for compact and incomplete map observations.
• Explored semantic-enhanced map representations and global descriptors for localization under sparse map observations.

Sparse-Map LiDAR Place Recognition

The work explores semantic-enhanced representations and global descriptors for localization under sparse map observations, with evaluation focused on sparse, incomplete, and deployment-oriented reference availability.