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

HEZAM ALBAQAMI 1, HAITIAN WANG 2, XINYU WANG 2, MUHAMMAD IBRAHIM 2, ZAINY M. MALAKAN 3, ABDULLAH M. ALGAMDI 1, MOHAMMED H. ALGHAMDI 4,5, AJMAL MIAN 2 · 2025

Read Urban 3D Change Detection Using LiDAR Sensor for HD Map Maintenance and Smart Mobility by HEZAM ALBAQAMI 1, HAITIAN WANG 2, XINYU WANG 2, MUHAMMAD IBRAHIM 2, ZAINY M. MALAKAN 3, ABDULLAH M. ALGAMDI 1, MOHAMMED H. ALGHAMDI 4,5, AJMAL MIAN 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-21112.txt.

Introduction

Hezam Albaqami leads seven co-authors on city scale three dimensional change detection from LiDAR scans, for planning, compliance, and map upkeep covering buildings and urban greenery alike. Standard surface differencing suffers vertical bias and viewpoint mismatch, while raw point cloud methods demand huge memory and perfect alignment.

Their uncertainty aware, object centric method aligns scans from different times with a multi resolution distribution transform and reasons about changed objects rather than raw points. Thin structures that voxel methods degrade are better preserved. Cities gain a route to fresh high definition maps without resurveying everything by hand.

Catalog record

CREATORHEZAM ALBAQAMI 1, HAITIAN WANG 2, XINYU WANG 2, MUHAMMAD IBRAHIM 2, ZAINY M. MALAKAN 3, ABDULLAH M. ALGAMDI 1, MOHAMMED H. ALGHAMDI 4,5, AJMAL MIAN 2
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-21112

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Urban 3D Change Detection Using LiDAR Sensor for HD Map Maintenance and Smart Mobility HEZAM ALBAQAMI 1, HAITIAN WANG 2, XINYU WANG 2, MUHAMMAD IBRAHIM 2, ZAINY M. MALAKAN 3, ABDULLAH M. ALGAMDI 1, MOHAMMED H. ALGHAMDI 4,5, AJMAL MIAN 2 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. The Subiaco 2025 HD LiDAR dataset and annotations are released at IEEE DataPort , and the implementation is available at GitHub . † † history: Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000. † † doi: 10.1109/ACCESS.2017.DOI † † address: Department of Computer Science and Artificial Intelligence, College of Computer Science and Engineering, University of Jeddah, Jeddah 21493, Saudi Arabia † † address: Department of Computer Science and Software Engineering, University of Western Australia, Perth, WA 6009, Australia † † address: Data Science Department, Umm Al-Qura University, Makkah 24382, Saudi Arabia. † † address: Department of Information and Technology Systems, College of Computer Science and Engineering, University of Jeddah, Jeddah 21493, Saudi Arabia. † † address: Department of Informatics and Computer Systems, College of Computer Science, King Khalid University, Abha, Saudi Arabia. † † titlenote: This work was funded by the University of Jeddah, Jeddah, Saudi Arabia, under grant No.(UJ-24-SUTU-1290). The authors thank the University of Jeddah for its technical and financial support. This project is a collaborative effort between the University of Jeddah and the University of Western Australia. † † corresponding: Corresponding author: Hezam Albaqami (e-mail: [email protected]). I Introduction Western Australia is advancing smart transportation, digital-twin, and autonomous driving deployments that depend on high-definition 3D city maps maintained at operational cadence [ 1 , 2 ] . Within this context, change detection denotes the object-level identification of additions, removals, and volumetric increases or decreases across bi-temporal LiDAR maps, enabling HD-map maintenance, construction and asset monitoring, and timely updates for localization and planning in autonomy stacks [ 3 , 4 , 5 , 6 , 7 ] . LiDAR sensing further supports this task by providing dense geometry that separates long-term structural changes from transient dynamics such as vehicles and pedestrians [ 8 , 9 , 10 ] . Subiaco lies immediately west of Perth’s CBD and functions as a key transport hub with two arterial roads linking the areas and substantial daily commuting, which makes it a representative and consequential testbed [ 11 , 12 ] . Addressing city-scale change detection in Subiaco is therefore directly aligned with Western Australia’s smart-city and autonomy objectives and motivates the problem tackled in this work [ 13 ] . Existing approaches to urban 3D change detection fall into height or digital surface model differencing, 2D image based change detection, and point based deep models [ 14 , 15 ] . Height differencing requires sub decimeter co registration and uniform sampling [ 16 , 17 ] . In dense street scenes small vertical biases and ground slope introduce systematic errors, curb lines and façade parallax generate false positives, and the output remains cell wise without object identity [ 18 , 16 ] . Image based change detection cannot resolve vertical geometry and suffers from viewpoint mismatch with ground LiDAR [ 19 , 20 ] . Point wise Siamese or triplet networks such as KPConv variants operate on cropped tiles with millions of points [ 21 , 22 , 23 ] . Runtime and memory grow with area, training assumes balanced labels, and predictions are per point rather than per object, which weakens Added and Removed reasoning when overlap is partial or occlusion changes sampling [ 24 , 25 ] . Voxel or bird’s eye encodings reduce memory but erode thin structures and narrow ground changes, which forces a trade off between coverage and fidelity [ 26 , 27 ] . Public benchmarks for 3D change detection cover limited extents or rely on aerial campaigns, and object level annotations across epochs are rare [ 28 , 29 ] . For Subiaco, publicly accessible LiDAR captures are single epoch and not paired at consistent specification, and no object level multi epoch ground truth exists [ 28 ] . Many pipelines also assume perfect pre alignment and do not propagate registration uncertainty into decision thresholds, which leads to spurious changes near loop junctions or under canopy [ 30 , 31 , 32 , 33 , 34 ] . Methods seldom enforce class consistent association between epochs, so split and merge cases remain unresolved and per class counts drift at city scale [ 35 , 36 , 29 ] . These constraints limit transfer to operational HD map maintenance over a dense transport hub and expose the need for a city scale, object centric and uncertainty aware formulation. Fig. 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). Our changing detection method is object centric and uncertainty aware, and it operates at city scale on bi-temporal LiDAR maps. An overview of the acquisition, mapping, and change-analysis stages is shown in Fig. 1 . It begins with global alignment of the two epochs using multi-resolution NDT followed by point-to-plane ICP, after which both maps are height normalized and a per-location level-of-detection is derived from registration covariance and surface roughness. This step replaces the usual perfect pre-alignment assumption and supplies calibrated thresholds that suppress spurious changes along loop junctions and under canopy. We then generate geometry-only proxies to seed cross-epoch associations and perform semantic and instance segmentation for ground, building, vegetation, and mobile classes on single-epoch geometry. Class-constrained bipartite matching refines correspondences while handling split and merge cases through an augmented assignment with dummies, which prevents drift

Read the complete text