Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

WHU-PCPR: A cross-platform heterogeneous point cloud dataset for place recognition in complex urban scenes

Xianghong Zou, Jianping Li, Yandi Yang, Weitong Wu, Yuan Wang, Qiegen Liu, Zhen Dong

TL;DR

WHU-PCPR tackles the lack of diversity in point-cloud place recognition by introducing a cross-platform, heterogeneous PCPR dataset that integrates MLS and portable helmet-mounted LiDAR data over 82.3 km and 60 months. The dataset enables a comprehensive benchmark, revealing substantial generalization gaps across scenes and sensor types, and the limited but beneficial impact of reranking in improving retrieval when paired with strong initial baselines. Through systematic evaluation of five retrieval methods and three reranking strategies, the work highlights the importance of domain adaptation, robust feature extraction, and geometry-based refinement for practical PCPR in complex urban environments. The work provides a valuable resource for advancing PCPR toward robust, cross-domain localization in real-world scenarios and sets a clear agenda for future research in generalization, scene changes, and viewpoint invariance.

Abstract

Point Cloud-based Place Recognition (PCPR) demonstrates considerable potential in applications such as autonomous driving, robot localization and navigation, and map update. In practical applications, point clouds used for place recognition are often acquired from different platforms and LiDARs across varying scene. However, existing PCPR datasets lack diversity in scenes, platforms, and sensors, which limits the effective development of related research. To address this gap, we establish WHU-PCPR, a cross-platform heterogeneous point cloud dataset designed for place recognition. The dataset differentiates itself from existing datasets through its distinctive characteristics: 1) cross-platform heterogeneous point clouds: collected from survey-grade vehicle-mounted Mobile Laser Scanning (MLS) systems and low-cost Portable helmet-mounted Laser Scanning (PLS) systems, each equipped with distinct mechanical and solid-state LiDAR sensors. 2) Complex localization scenes: encompassing real-time and long-term changes in both urban and campus road scenes. 3) Large-scale spatial coverage: featuring 82.3 km of trajectory over a 60-month period and an unrepeated route of approximately 30 km. Based on WHU-PCPR, we conduct extensive evaluation and in-depth analysis of several representative PCPR methods, and provide a concise discussion of key challenges and future research directions. The dataset and benchmark code are available at https://github.com/zouxianghong/WHU-PCPR.

WHU-PCPR: A cross-platform heterogeneous point cloud dataset for place recognition in complex urban scenes

TL;DR

WHU-PCPR tackles the lack of diversity in point-cloud place recognition by introducing a cross-platform, heterogeneous PCPR dataset that integrates MLS and portable helmet-mounted LiDAR data over 82.3 km and 60 months. The dataset enables a comprehensive benchmark, revealing substantial generalization gaps across scenes and sensor types, and the limited but beneficial impact of reranking in improving retrieval when paired with strong initial baselines. Through systematic evaluation of five retrieval methods and three reranking strategies, the work highlights the importance of domain adaptation, robust feature extraction, and geometry-based refinement for practical PCPR in complex urban environments. The work provides a valuable resource for advancing PCPR toward robust, cross-domain localization in real-world scenarios and sets a clear agenda for future research in generalization, scene changes, and viewpoint invariance.

Abstract

Point Cloud-based Place Recognition (PCPR) demonstrates considerable potential in applications such as autonomous driving, robot localization and navigation, and map update. In practical applications, point clouds used for place recognition are often acquired from different platforms and LiDARs across varying scene. However, existing PCPR datasets lack diversity in scenes, platforms, and sensors, which limits the effective development of related research. To address this gap, we establish WHU-PCPR, a cross-platform heterogeneous point cloud dataset designed for place recognition. The dataset differentiates itself from existing datasets through its distinctive characteristics: 1) cross-platform heterogeneous point clouds: collected from survey-grade vehicle-mounted Mobile Laser Scanning (MLS) systems and low-cost Portable helmet-mounted Laser Scanning (PLS) systems, each equipped with distinct mechanical and solid-state LiDAR sensors. 2) Complex localization scenes: encompassing real-time and long-term changes in both urban and campus road scenes. 3) Large-scale spatial coverage: featuring 82.3 km of trajectory over a 60-month period and an unrepeated route of approximately 30 km. Based on WHU-PCPR, we conduct extensive evaluation and in-depth analysis of several representative PCPR methods, and provide a concise discussion of key challenges and future research directions. The dataset and benchmark code are available at https://github.com/zouxianghong/WHU-PCPR.
Paper Structure (16 sections, 8 figures, 4 tables)

This paper contains 16 sections, 8 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Public PCPR datasets
  4. PCPR methods
  5. Construction of WHU-PCPR dataset
  6. Data collection
  7. Data preprocessing
  8. Characteristics of WHU-PCPR
  9. Benchmark
  10. Experiment setting
  11. Quantitative evaluation for PCPR
  12. Retrieval results
  13. Reranking results
  14. Challenges and research directions
  15. Conclusion
  16. ...and 1 more sections

Figures (8)

  • Figure 1: Overview of WHU-PCPR dataset. A, B, and C represent three typical scenes.
  • Figure 2: Cloud to cloud distance in WHU-PCPR. (a) WHU 1&2 (CS College), (b) WHU 1&2 (Info Campus), (c) Hankou 1&2 (Zhongshan Park), (d) Hankou 1&2 (Jiefang Road 1), (e) WHU 1&3 (CS College), (f) WHU 1&3 (Info Campus), (g) Hankou 1&3 (Zhongshan Park), (h) Hankou 1&3 (Jiefang Road 1). A, B, C, and D are the positional errors of manually selected corresponding points (gray/blue/red: phase 1/2/3).
  • Figure 3: Characteristics of WHU-PCPR.
  • Figure 4: Recall curves of retrieval baselines on WHU-PCPR. (a) Hankou 1&2, (b) Hankou 1&3, (c) Hankou 2&3, (d) WHU 1&2, (e) WHU 1&3, (f) WHU 2&3.
  • Figure 5: Retrieval results of various baselines with different viewpoints. (a) $R@1$ on Hankou 1&2, (b) $R@1\%$ on Hankou 1&2, (c) $R@1$ on WHU 1&2, (d) $R@1\%$ on WHU 1&2.
  • ...and 3 more figures