Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Urban 3D Change Detection Using LiDAR Sensor for HD Map Maintenance and Smart Mobility

Hezam Albagami, Haitian Wang, Xinyu Wang, Muhammad Ibrahim, Zainy M. Malakan, Abdullah M. Alqamdi, Mohammed H. Alghamdi, Ajmal Mian

TL;DR

Problem: robust object-level 3D change detection for city-scale HD maps using bi-temporal LiDAR. Approach: an object-centric, uncertainty-aware registration and change analysis pipeline that derives per-object correspondences and gates decisions by local uncertainty. Findings: on Subiaco 2025 vs 2023, the method achieves 95.2% accuracy, 90.4% macro F1, and 82.6% macro IoU, outperforming state-of-the-art baselines. Significance: enables reliable HD-map maintenance and autonomous mobility with memory-bounded, tiled processing and a public Subiaco 2025 dataset.

Abstract

High-definition 3D city maps underpin smart transportation, digital twins, and autonomous driving, where object level change detection across bi temporal LiDAR enables HD map maintenance, construction monitoring, and reliable localization. Classical DSM differencing and image based methods are sensitive to small vertical bias, ground slope, and viewpoint mismatch and yield cellwise outputs without object identity. Point based neural models and voxel encodings demand large memory, assume near perfect pre alignment, degrade thin structures, and seldom enforce class consistent association, which leaves split or merge cases unresolved and ignores uncertainty. We propose an object centric, uncertainty aware pipeline for city scale LiDAR that aligns epochs with multi resolution NDT followed by point to plane ICP, normalizes height, and derives a per location level of detection from registration covariance and surface roughness to calibrate decisions and suppress spurious changes. Geometry only proxies seed cross epoch associations that are refined by semantic and instance segmentation and a class constrained bipartite assignment with augmented dummies to handle splits and merges while preserving per class counts. Tiled processing bounds memory without eroding narrow ground changes, and instance level decisions combine 3D overlap, normal direction displacement, and height and volume differences with a histogram distance, all gated by the local level of detection to remain stable under partial overlap and sampling variation. On 15 representative Subiaco blocks the method attains 95.2% accuracy, 90.4% mF1, and 82.6% mIoU, exceeding Triplet KPConv by 0.2 percentage points in accuracy, 0.2 in mF1, and 0.8 in mIoU, with the largest gain on Decreased where IoU reaches 74.8% and improves by 7.6 points.

Urban 3D Change Detection Using LiDAR Sensor for HD Map Maintenance and Smart Mobility

TL;DR

Problem: robust object-level 3D change detection for city-scale HD maps using bi-temporal LiDAR. Approach: an object-centric, uncertainty-aware registration and change analysis pipeline that derives per-object correspondences and gates decisions by local uncertainty. Findings: on Subiaco 2025 vs 2023, the method achieves 95.2% accuracy, 90.4% macro F1, and 82.6% macro IoU, outperforming state-of-the-art baselines. Significance: enables reliable HD-map maintenance and autonomous mobility with memory-bounded, tiled processing and a public Subiaco 2025 dataset.

Abstract

High-definition 3D city maps underpin smart transportation, digital twins, and autonomous driving, where object level change detection across bi temporal LiDAR enables HD map maintenance, construction monitoring, and reliable localization. Classical DSM differencing and image based methods are sensitive to small vertical bias, ground slope, and viewpoint mismatch and yield cellwise outputs without object identity. Point based neural models and voxel encodings demand large memory, assume near perfect pre alignment, degrade thin structures, and seldom enforce class consistent association, which leaves split or merge cases unresolved and ignores uncertainty. We propose an object centric, uncertainty aware pipeline for city scale LiDAR that aligns epochs with multi resolution NDT followed by point to plane ICP, normalizes height, and derives a per location level of detection from registration covariance and surface roughness to calibrate decisions and suppress spurious changes. Geometry only proxies seed cross epoch associations that are refined by semantic and instance segmentation and a class constrained bipartite assignment with augmented dummies to handle splits and merges while preserving per class counts. Tiled processing bounds memory without eroding narrow ground changes, and instance level decisions combine 3D overlap, normal direction displacement, and height and volume differences with a histogram distance, all gated by the local level of detection to remain stable under partial overlap and sampling variation. On 15 representative Subiaco blocks the method attains 95.2% accuracy, 90.4% mF1, and 82.6% mIoU, exceeding Triplet KPConv by 0.2 percentage points in accuracy, 0.2 in mF1, and 0.8 in mIoU, with the largest gain on Decreased where IoU reaches 74.8% and improves by 7.6 points.
Paper Structure (14 sections, 8 equations, 7 figures, 2 tables)

This paper contains 14 sections, 8 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Subiaco 2025 3D LiDAR Map Construction
  3. Survey Plan and Sensor Setup
  4. Acquisition Summary
  5. Map Generation
  6. Methodology
  7. Geometric Registration and Object-Level Correspondence
  8. Semantic and Instance Segmentation
  9. Change Analysis
  10. Experiments
  11. Subiaco 2025 3D Map Construction Result
  12. Qualitative Change Detection on Representative Blocks
  13. Quantitative Evaluation on City Blocks and Method Comparison
  14. Conclusion

Figures (7)

  • Figure 1: Overview of the Subiaco city-scale 3D change-detection pipeline. Top: acquisition/mapping (soft GNSS sync, ground segmentation, sequence construction, NDT $\rightarrow$ ICP). Bottom: analysis (cross-epoch association, semantic/object segmentation, and uncertainty-gated instance labels).
  • Figure 2: Subiaco driving loops for the 2025 campaign. Colored polylines denote twenty overlapping loops that provide repeated viewpoints across arterials, residential blocks, and intersections.
  • Figure 3: Sensor rig and installation. The system comprises an Ouster OS1-128 LiDAR, a dual-antenna RTK GNSS with MEMS IMU, and a logging unit mounted on a reinforced roof plate with a measured antenna baseline.
  • Figure 4: Object-centric, uncertainty-aware pipeline. Left-to-right: registration and correspondence, semantic/instance segmentation with class-consistent association, change analysis with LoD gating and confidence fusion, and the instance-level result.
  • Figure 5: Examples of the four semantic categories: (a) vegetation, (b) mobile object, (c) building, and (d) other object (e.g., pole or ground fragment).
  • ...and 2 more figures