ResLPR: A LiDAR Data Restoration Network and Benchmark for Robust Place Recognition Against Weather Corruptions

2025-03-16

AI Summary

Existing LiDAR-based place recognition (LPR) methods suffer severe performance degradation under adverse weather conditions (e.g., rain, snow, fog) due to weather-induced LiDAR point cloud degradation. To address this, we propose ResLPRNet, a lightweight wavelet-transform-driven LiDAR data restoration network, and introduce ResLPR—the first LPR benchmark encompassing diverse weather corruptions. ResLPRNet features a novel plug-and-play wavelet encoder-decoder architecture for end-to-end modeling of point cloud distortions. Furthermore, we present WeatherKITTI and WeatherNCLT—two synthetic data generation frameworks that bridge the critical gap in adverse-weather LPR training and evaluation data. Extensive experiments demonstrate that ResLPRNet consistently improves recall@1 by over 25% across multiple state-of-the-art LPR models, with negligible computational overhead (<3% FLOPs increase), significantly enhancing robustness and practicality for LPR under rain, snow, and fog.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Natural Language Processing: Safety and Robustness
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

LiDAR-based place recognition (LPR) is a key component for autonomous driving, and its resilience to environmental corruption is critical for safety in high-stakes applications. While state-of-the-art (SOTA) LPR methods perform well in clean weather, they still struggle with weather-induced corruption commonly encountered in driving scenarios. To tackle this, we propose ResLPRNet, a novel LiDAR data restoration network that largely enhances LPR performance under adverse weather by restoring corrupted LiDAR scans using a wavelet transform-based network. ResLPRNet is efficient, lightweight and can be integrated plug-and-play with pretrained LPR models without substantial additional computational cost. Given the lack of LPR datasets under adverse weather, we introduce ResLPR, a novel benchmark that examines SOTA LPR methods under a wide range of LiDAR distortions induced by severe snow, fog, and rain conditions. Experiments on our proposed WeatherKITTI and WeatherNCLT datasets demonstrate the resilience and notable gains achieved by using our restoration method with multiple LPR approaches in challenging weather scenarios. Our code and benchmark are publicly available here: https://github.com/nubot-nudt/ResLPR.

Problem

• Enhances LiDAR-based place recognition in adverse weather
• Proposes a lightweight network for restoring corrupted LiDAR data
• Introduces a benchmark for testing LPR methods under severe weather

Innovation

• Wavelet transform-based LiDAR data restoration
• Plug-and-play integration with pretrained models
• Novel benchmark for adverse weather conditions