Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
March 29, 2026Remote SensingOpen Access

Long-Tail Learning for Three-Dimensional Pavement Distress Segmentation Using Point Clouds Reconstructed from a Consumer Camera

View Full Paper
Ask AI
Bookmark
Share

Authors

PCPengjian ChengHarbin Institute of TechnologyJYJunyan YiHarbin Institute of TechnologyZPZhongshi PeiHarbin Institute of Technology

Discussion

Loading...

Member takes

Implication

Demonstrates effective 3D distress segmentation in pavements, suggesting new methods for monitoring and evaluation.

Key Points

  • The aim is to improve 3D pavement distress assessment using consumer-grade cameras and address class imbalances in segmentation.
  • Utilized a consumer-grade action camera for data collection.
  • Developed a 3D point cloud dataset of pavements.
  • Implemented a long-tail class imbalance mitigation strategy with adaptive re-sampling and weighted fusion loss.
  • Designed PointPaveSeg, a specialized point cloud processing network with decoupled geometric and semantic features.
  • Conducted field evaluations of segmented point clouds against manual measurements.
  • Achieved a mean Intersection over Union (mIoU) of 78.45%.
  • Obtained an accuracy rate of 95.43%.
  • Showed high consistency with manual measurements during field evaluations.
  • Validated the method's applicability for real-world pavement monitoring and maintenance systems.

Cite This Study

Cheng et al. (2026) studied this question.

synapsesocial.com/papers/69c8c2d1de0f0f753b39d473https://doi.org/10.3390/rs18071008
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1PAN: Improved PointNet++ for Pavement Crack Information Extraction2024 · 10 citations
  2. 2Deep Long-Tailed Learning: A Survey2023 · 684 citations
  3. 3LiDAR-Based Road Cracking Detection: Machine Learning Comparison, Intensity Normalization, and Open-Source WebGIS for Infrastructure Maintenance2025 · 13 citations
  4. 4GCN-Based Pavement Crack Detection Using Mobile LiDAR Point Clouds2021 · 86 citations
  5. 5PCT: Point cloud transformer2021 · 1,908 citations