3D LULC classification using multispectral LiDAR and deep learning: current and prospective schemes

Aldino Rizaldy; Fabio Remondino; Gottfried Mandlburger; Juha Hyypp\"a; Markus Hollaus; Narges Takhtkeshha

arxiv: 2605.22328 · v1 · pith:ODEBAMEInew · submitted 2026-05-21 · 💻 cs.CV

3D LULC classification using multispectral LiDAR and deep learning: current and prospective schemes

Narges Takhtkeshha , Aldino Rizaldy , Markus Hollaus , Juha Hyypp\"a , Fabio Remondino , Gottfried Mandlburger This is my paper

Pith reviewed 2026-05-22 06:45 UTC · model grok-4.3

classification 💻 cs.CV

keywords multispectral LiDAR3D LULC classificationpoint cloud semantic segmentationdeep learningNMCA schemesLoosdorf-MSL datasetPoint Transformer V3

0 comments

The pith

Multispectral LiDAR with deep learning reaches 79.4 percent mIoU on new national-standard 3D land cover schemes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces L1 and L2 classification schemes aligned with National Mapping and Cadastral Agencies standards along with the Loosdorf-MSL benchmark dataset of multispectral LiDAR point clouds. It tests seven deep learning models for semantic segmentation of these 3D data and finds that Point Transformer V3 performs best when given both geometric coordinates and reflectance values from 532 nm and 1064 nm lasers. Multispectral inputs raise mean intersection over union by 1.1 points at the coarse 8-class level and by 7.8 points at the fine 20-class level compared with geometry alone. These resources address the shortage of public urban datasets that match official mapping schemes and demonstrate that reflectance helps distinguish materials at higher semantic detail. The work therefore supplies concrete tools and evidence for consistent national 3D LULC mapping.

Core claim

The paper presents the Loosdorf-MSL multispectral LiDAR dataset and two NMCA-aligned LULC schemes (L1 with 8 classes and L2 with 20 classes) for 3D point cloud semantic segmentation. Point Transformer V3 achieves the highest scores of 79.4 percent mIoU on L1 and 58.9 percent mIoU on L2 when both wavelengths are used. Ablation tests show that adding the spectral reflectance channels improves results over geometry-only inputs, with larger gains at the finer L2 level where material discrimination matters most.

What carries the argument

Point Transformer V3 neural network applied to dual-wavelength multispectral LiDAR point clouds that combine 3D geometry with reflectance at 532 nm and 1064 nm under the L1 and L2 NMCA-aligned schemes

If this is right

National mapping agencies gain ready-to-use L1 and L2 schemes that match their existing classification needs for 3D data.
Reflectance from two laser wavelengths supplies additional cues that raise accuracy on fine material distinctions at the 20-class level.
Deep learning models trained on this benchmark can serve as starting points for operational 3D LULC production pipelines.
The public Loosdorf-MSL dataset enables direct comparison of future algorithms against the reported Point Transformer V3 baseline.
Official LULC schemes can move toward greater semantic detail once spectral information is routinely available from LiDAR.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the spectral gains hold across varied climates and building stocks, agencies may shift procurement toward multispectral rather than single-wavelength LiDAR systems.
The same point-cloud models could be extended to fuse aerial imagery or other sensors for even richer labels without starting from scratch.
Creation of similar benchmark datasets in other countries would test whether the L1 and L2 schemes require only minor translation or need substantial redesign.
Higher-resolution or additional wavelength channels might further improve discrimination of vegetation subtypes or construction materials at the L2 level.

Load-bearing premise

The Loosdorf-MSL dataset from one region is representative of typical urban and suburban settings so that the L1 and L2 schemes and reported model performances transfer directly to other National Mapping and Cadastral Agencies without major local changes.

What would settle it

Running the same seven models and L1/L2 schemes on a new multispectral LiDAR dataset collected from a different city or country and checking whether the 79.4 percent and 58.9 percent mIoU figures plus the 1.1 and 7.8 point spectral gains still appear.

Figures

Figures reproduced from arXiv: 2605.22328 by Aldino Rizaldy, Fabio Remondino, Gottfried Mandlburger, Juha Hyypp\"a, Markus Hollaus, Narges Takhtkeshha.

**Figure 8.** Figure 8: Spectral histogram of different LULC classes across different laser wavelengths. Reflectance values are scaled between 0 and 1 for visualization purposes. 3.2. Importance of the Loosdorf-MSL dataset and comparison with existing datasets To date, considerable efforts exist in preparing ALS/ULS (UAV-LiDAR)-based LULC classification benchmark datasets, specifically to facilitate the training, evaluation, and … view at source ↗

read the original abstract

Land Use Land Cover (LULC) classification is essential for national 3D mapping, geospatial analysis, and sustainable planning. Multispectral (MS) LiDAR provides synchronized spatial-spectral information, and deep learning (DL) enables 3D point cloud semantic segmentation; however, adoption is limited by the lack of publicly available urban and suburban MS LiDAR datasets aligned with National Mapping and Cadastral Agencies (NMCAs) classification schemes. This study addresses these gaps by introducing L1 and L2 NMCA-aligned LULC classification schemes and a new benchmark MS LiDAR dataset. We evaluate seven state-of-the-art DL models and perform spectral ablation studies at both levels of detail. Results show that Point Transformer V3 achieves the best performance, with mIoU of 79.4% (L1, 8 classes) and 58.9% (L2, 20 classes) using a dual-wavelength LiDAR system (532 nm and 1064 nm). Ablation results show that multispectral information improves performance over geometry-only inputs, with gains of 1.1 percentage points at L1 and 7.8 points at L2. These results highlight the value of LiDAR reflectance for fine-grained material discrimination and support the evolution of NMCA LULC schemes toward higher semantic detail. The Loosdorf-MSL dataset contributes a new benchmark for consistent national and international LULC mapping.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

New dataset and NMCA-aligned schemes are the concrete addition here, but single-site results limit how far the adoption claims can stretch.

read the letter

This paper's main contribution is the Loosdorf-MSL multispectral LiDAR dataset plus the L1 and L2 classification schemes aligned with national mapping agency standards. They run seven deep learning models on 3D point cloud segmentation, report Point Transformer V3 at the top with 79.4% mIoU on the 8-class level and 58.9% on the 20-class level, and show that the dual-wavelength spectral channels add 1.1 points at L1 and 7.8 points at L2 over geometry-only inputs. The ablation work is straightforward and useful for showing where reflectance helps with material distinctions at finer granularity. The dataset release and the explicit tie to NMCA schemes fill a documented gap in public benchmarks, which is the part that actually moves the practical side forward. The soft spot is the single test site. The stress-test point holds: one location does not automatically demonstrate that the schemes or the performance numbers will transfer to the range of urban and suburban conditions that national agencies handle. Without broader sampling or clear discussion of how class distributions and site characteristics compare to other regions, the push toward direct adoption rests on an assumption that needs more support. Dataset size, balance, and training protocol details matter for judging the numbers, and any full version should make those explicit. This is for people doing LiDAR semantic segmentation or working on operational 3D mapping pipelines. A reader who needs a new benchmark dataset or wants to see how current models handle multispectral inputs will get direct value from the comparisons and the released data. It deserves peer review because the dataset and schemes are tangible items others can use and test, even if the generalization questions require attention in revision.

Referee Report

1 major / 2 minor

Summary. The manuscript introduces L1 (8-class) and L2 (20-class) LULC classification schemes aligned with National Mapping and Cadastral Agencies (NMCAs), presents the new Loosdorf-MSL multispectral LiDAR benchmark dataset, and benchmarks seven deep learning models for 3D point cloud semantic segmentation. Point Transformer V3 achieves the highest mIoU of 79.4% (L1) and 58.9% (L2) with dual-wavelength (532 nm + 1064 nm) inputs; ablation studies show multispectral reflectance improves over geometry-only baselines by 1.1 pp at L1 and 7.8 pp at L2. The work positions these contributions as filling gaps in public datasets and supporting evolution toward higher-detail NMCA schemes.

Significance. If the reported performance numbers and ablation gains hold under broader validation, the paper supplies a useful new public benchmark dataset and concrete empirical evidence that multispectral LiDAR reflectance aids fine-grained material discrimination in 3D urban/suburban scenes. The two-level NMCA-aligned taxonomy and systematic comparison of recent point-cloud transformers constitute a practical step toward standardized 3D LULC mapping.

major comments (1)

[Abstract] Abstract: The claim that the L1/L2 schemes can be directly adopted by NMCAs without major local adaptations rests on the assumption that the single-site Loosdorf-MSL dataset is representative of the range of urban/suburban geometries, roof types, vegetation, and class distributions encountered nationally. No multi-site or cross-region validation is presented, so both the absolute mIoU figures and the magnitude of the multispectral gains may not generalize.

minor comments (2)

[Abstract] Abstract and methods: Dataset size (number of points or scenes), class distribution statistics, train/val/test splits, and training protocols are not summarized, limiting the reader's ability to assess the reliability of the reported mIoU values and ablation deltas.
Ensure that implementation details for the seven evaluated models (hyperparameters, data augmentation, loss functions) are provided with sufficient precision for reproducibility, ideally with a link to code or configuration files.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback. We address the major comment below and describe the revisions we will implement.

read point-by-point responses

Referee: [Abstract] Abstract: The claim that the L1/L2 schemes can be directly adopted by NMCAs without major local adaptations rests on the assumption that the single-site Loosdorf-MSL dataset is representative of the range of urban/suburban geometries, roof types, vegetation, and class distributions encountered nationally. No multi-site or cross-region validation is presented, so both the absolute mIoU figures and the magnitude of the multispectral gains may not generalize.

Authors: We agree with the referee that the single-site character of the Loosdorf-MSL dataset precludes strong claims of direct national adoption without local adaptations. The manuscript presents no multi-site or cross-region experiments, so the reported mIoU values and the size of the multispectral gains cannot be assumed to hold elsewhere. We will revise the abstract to eliminate any implication of immediate NMCA-wide applicability. The new wording will describe the L1/L2 schemes as NMCA-aligned taxonomies and the dataset as a public benchmark for method development, while explicitly noting that further multi-site validation is required before broader deployment. We will also add a short paragraph in the discussion section that states this limitation and outlines the need for future cross-region studies. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical results on newly introduced dataset

full rationale

The paper introduces L1/L2 NMCA-aligned classification schemes and the Loosdorf-MSL dataset, then reports mIoU and ablation gains from standard training and evaluation of existing deep learning models (Point Transformer V3 and six others) on held-out test splits. No equations, parameters, or premises are defined in terms of the reported outcomes; the performance numbers are produced by conventional supervised learning pipelines rather than by construction or self-referential fitting. No load-bearing self-citations, uniqueness theorems, or ansatzes appear in the core claims. The derivation chain is therefore self-contained empirical measurement.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

The work rests on standard assumptions about deep learning applicability to point clouds and the utility of multispectral LiDAR for material discrimination. It introduces new classification schemes and a dataset without upstream independent validation.

free parameters (1)

Deep learning model hyperparameters
Typical tuning of learning rates, batch sizes, and architectures in the seven evaluated models to achieve reported mIoU values.

axioms (1)

domain assumption Multispectral LiDAR provides synchronized spatial-spectral information that aids semantic segmentation beyond geometry alone.
Invoked to justify the value of dual-wavelength data in the ablation studies.

invented entities (2)

L1 and L2 NMCA-aligned LULC classification schemes no independent evidence
purpose: Provide standardized levels of detail for 3D national mapping.
Newly proposed schemes to address lack of aligned public datasets.
Loosdorf-MSL dataset no independent evidence
purpose: Benchmark multispectral LiDAR data for LULC classification research.
New dataset contributed to enable consistent national and international mapping.

pith-pipeline@v0.9.0 · 5823 in / 1493 out tokens · 58802 ms · 2026-05-22T06:45:45.611891+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Point Transformer V3 achieves the best performance, with mIoU of 79.4% (L1, 8 classes) and 58.9% (L2, 20 classes) using a dual-wavelength LiDAR system (532 nm and 1064 nm).
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Ablation results show that multispectral information improves performance over geometry-only inputs, with gains of 1.1 percentage points at L1 and 7.8 points at L2.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

32 extracted references · 32 canonical work pages

[1]

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-2-2024

ESTATE: A Large Dataset of Under -Represented Urban Objects for 3D Point Cloud Classification. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-2-2024. https://doi.org/10.5194/isprs-archives-XLVIII-2-2024-25-2024. Bayrak, O. C., Remondino, F., & Uzar, M.,

work page doi:10.5194/isprs-archives-xlviii-2-2024-25-2024 2024
[2]

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII -1/W3-2023

A new dataset and methodology for urban -scale 3D point cloud classification. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII -1/W3-2023. https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-1-2023. Carós, M., Just, A., Seguí, S., & Vitrià, J.,

work page doi:10.5194/isprs-archives-xlviii-1-w3-2023-1-2023 2023
[3]

ISPRS Open Journal of Photogrammetry and Remote Sensing, 19, 100119

Enhancing point cloud semantic segmentation via scalable domain adaptation with LoRA-enabled PointNet++. ISPRS Open Journal of Photogrammetry and Remote Sensing, 19, 100119. https://doi.org/10.1016/j.ophoto.2026.100119. Chen, Z.,

work page doi:10.1016/j.ophoto.2026.100119 2026
[4]

In: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Geospatial Week 2019, Enschede, NL

Hybrid orientation of airborne lidar point clouds and aerial images. In: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Geospatial Week 2019, Enschede, NL. Graham, B., Engelcke, M., & Maaten, L.,

work page 2019
[5]

3D semantic segmentation with submanifold sparse convolutional networks. Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 9224–9232. https://doi.ieeecomputersociety.org/10.1109/CVPR.2018.00961. Han, X., Liu, C., Dong, Z., & Yang, B.,

work page doi:10.1109/cvpr.2018.00961 2018
[6]

In: Proceedings of the IEEE/CVF Conference on Computer 25 Vision and Pattern Recognition, pp

RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 11105–11114. https://doi.org/10.1109/CVPR42600.2020.01112. Jing, Z., Guan, H., Zhao, P., Li, D., Y u, Y ., Zang, Y ., Wang, H., & Li, J.,

work page doi:10.1109/cvpr42600.2020.01112 2020
[7]

IEEE Geosci

Toward Hyperspectral Lidar: Measurement of Spectral Backscatter Intensity With a Supercontinuum Laser Source. IEEE Geosci. Remote Sens. Lett. 4, 211–215. https://doi.org/10.1109/LGRS.2006.888848. Karila, K., Matikainen, L., Litkey, P., Hyyppä, J., & Puttonen, E.,

work page doi:10.1109/lgrs.2006.888848 2006
[8]

International Journal of Remote Sensing, 40, 1 –19

The effect of seasonal variation on automated land cover mapping from multispectral airborne laser scanning data. International Journal of Remote Sensing, 40, 1 –19. https://doi.org/10.1080/01431161.2018.1528023. Kölle, M., Laupheimer, D., Schmohl, S., Haala, N., Rottensteiner, F., & Wegner, J.,

work page doi:10.1080/01431161.2018.1528023 2018
[9]

ISPRS Open Journal of Photogrammetry and Remote Sensing, 1, 100001

The Hessigheim 3D (H3D) benchmark on semantic segmentation of high -resolution 3D point clouds and textured meshes from UA V LiDAR and Multi -View-Stereo. ISPRS Open Journal of Photogrammetry and Remote Sensing, 1, 100001. https://doi.org/10.1016/j.ophoto.2021.100001. Li, D., Shen, X., Guan, H., Yu, Y ., Wang, H., Zhang, G., & Li, J.,

work page doi:10.1016/j.ophoto.2021.100001 2021
[10]

AGFP-Net: Attentive geometric feature pyramid network for land cover classification using airborne multispectral LiDAR data. Int. J. Appl. Earth Obs. Geoinf., 108, 102723. https://doi.org/10.1016/j.jag.2022.102723. Liao, Y ., Xie, J., & Geiger, A.,

work page doi:10.1016/j.jag.2022.102723 2022
[11]

IEEE Transactions on Pattern Analysis and Machine Intelligence

KITTI 360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3D. IEEE Transactions on Pattern Analysis and Machine Intelligence. https://doi.org/10.1109/TPAMI.2022.3179507. Mandlburger, G., Otepka, J., Wagner, W., Karel, W., & Pfeifer, N.,

work page doi:10.1109/tpami.2022.3179507 2022
[12]

https://www.researchgate.net/publication/237437116

Orientation and processing of airborne laser scanning data (OPALS) - Concept and first results of a comprehensive ALS software. https://www.researchgate.net/publication/237437116. Mikael Reichler, Josef Taher, Petri Manninen, Harri Kaartinen, Juha Hyyppä, & Antero Kukko,

work page arXiv
[13]

ISPRS Open Journal of Photogrammetry and Remote Sensing, 12, 100061

Semantic segmentation of raw multispectral laser scanning data from urban environments with deep neural networks. ISPRS Open Journal of Photogrammetry and Remote Sensing, 12, 100061. https://doi.org/10.1016/j.ophoto.2024.100061. Morsy, S., Shaker, A., & El-Rabbany, A.,

work page doi:10.1016/j.ophoto.2024.100061 2024
[14]

Nachtergaele, F

https://doi.org/10.3390/s17050958. Nachtergaele, F. O. F., & Licona-Manzur, C.,

work page doi:10.3390/s17050958
[15]

National Land Survey of Finland, (n.d.)

The Land Degradation Assessment in Drylands (LADA) Project: Reflections on Indicators for Land Degradation Assessment. National Land Survey of Finland, (n.d.). Laser scanning data 5 p. National Land Survey of Finland. https://www.maanmittauslaitos.fi/en/maps-and-spatial-data/datasets-and-interfaces/product-descriptions/laser-scanning-data-5-p (accessed Oc...

work page 2025
[16]

Land -cover classification of multispectral LiDAR data using CNN with optimized hyper -parameters. ISPRS J. Photogramm. Remote Sens., 166, 241 –254. https://doi.org/10.1016/j.isprsjprs.2020.05.022. Pan, S., Guan, H., Yu, Y ., Li, J., & Peng, D.,

work page doi:10.1016/j.isprsjprs.2020.05.022 2020
[17]

Qi, C.R., Su, H., Mo, K., & Guibas, L.J., 2017a

https://doi.org/10.1117/12.849641. Qi, C.R., Su, H., Mo, K., & Guibas, L.J., 2017a. PointNet: Deep learning on point sets for 3D classification and segmentation . Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 652–660. https://doi.org/10.1109/CVPR.2017.16. Qi, C.R., Yi, L., Su, H., Guibas, L.J., 2017b: PointNet++: Deep hierarchical feature learni...

work page doi:10.1117/12.849641 2017
[18]

Multi-Kernel Graph Structure Learning for Multispectral Point Cloud Classification. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., PP, 1–15. https://doi.org/10.1109/JSTARS.2024.3368472. Riveiro, B., Lindenbergh, R.,

work page doi:10.1109/jstars.2024.3368472 2024
[19]

ISPRS Journal of Photogrammetry and Remote Sensing, 236, 569–605

Benchmarking individual tree segmentation using multispectral airborne laser scanning data: The FGI -EMIT dataset. ISPRS Journal of Photogrammetry and Remote Sensing, 236, 569–605. https://doi.org/10.1016/j.isprsjprs.2026.04.021. Ruoppa, L., Oinonen, O., Taher, J., Lehtomäki, M., Takhtkeshha, N., Kukko, A., Kaartinen, H., Hyyppä, J.,

work page doi:10.1016/j.isprsjprs.2026.04.021 2026
[20]

Unsupervised deep learning for semantic segmentation of multispectral LiDAR forest point clouds. ISPRS J. Photogramm. Remote Sens. 228, 694–722. https://doi.org/10.1016/j.isprsjprs.2025.07.038. Saurabh Prasad, Bertrand Le Saux, Naoto Yokoya, & Ronny Hansch,

work page doi:10.1016/j.isprsjprs.2025.07.038 2025
[21]

IEEE Dataport

2018 IEEE GRSS Data Fusion Challenge – Fusion of Multispectral LiDAR and Hyperspectral Data. IEEE Dataport. https://doi.org/10.21227/jnh9-nz89. Serna, A., Marcotegui, B., Goulette, F., & Deschaud, J. -E.,

work page doi:10.21227/jnh9-nz89 2018
[22]

ICPRAM 2014, 1–4

Paris-rue-Madame database: a 3D mobile laser scanner dataset for benchmarking urban detection, segmentation and classification methods. ICPRAM 2014, 1–4. Taher, J., Hakala, T., Jaakkola, A., Hyyti, H., Kukko, A., Manninen, P., Maanpää, J., Hyyppä, J.,

work page 2014
[23]

Taher, J., Hyyppä, E., Hyyppä, M., Salolahti, K., Y u, X., Matikainen, L., Kukko, A., Lehtomäki, M., Kaartinen, H., Thurachen, S., et al.,

https://doi.org/10.3390/s22155759. Taher, J., Hyyppä, E., Hyyppä, M., Salolahti, K., Y u, X., Matikainen, L., Kukko, A., Lehtomäki, M., Kaartinen, H., Thurachen, S., et al.,

work page doi:10.3390/s22155759
[24]

Multispectral airborne laser scanning for tree species classification: A benchmark of machine learning and deep learning algorithms. ISPRS J. Photogramm. Remote Sens. 233, 278–309. https://doi.org/10.1016/j.isprsjprs.2026.01.031. Takhtkeshha, N., Bayrak, O.C., Mandlburger, G., Remondino, F., Kukko, A., Hyyppä, J.,2024a. Automatic annotation of 3D multispe...

work page doi:10.1016/j.isprsjprs.2026.01.031 2026
[25]

Takhtkeshha, N., Mazzacca, G., Remondino, F., Hyyppä, J., & Mandlburger, G., 2025b

https://doi.org/10.3390/s24051669. Takhtkeshha, N., Mazzacca, G., Remondino, F., Hyyppä, J., & Mandlburger, G., 2025b. Multispectral LiDAR data for extracting tree points in urban and suburban areas. https://doi.org/10.48550/arXiv.2508.19881. Thomas, H., Qi, C. R., Deschaud, J. -E., Marcotegui, B., Goulette, F., Guibas, L.,

work page doi:10.3390/s24051669
[26]

PoseNet: A convolutional network for real-time 6-dof camera relocalization,

KPConv: Flexible and deformable convolution for point clouds. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, South Korea, pp. 6410–6419. https://doi.org/10.1109/ICCV .2019.00651. Thomas, H., Tsai, Y .-H.H., Barfoot, T.D., Zhang, J.,

work page doi:10.1109/iccv 2019
[27]

The Visual Computer, 40(11), 8287–8329

Deep learning based computer vision under the prism of 3D point clouds: a systematic review. The Visual Computer, 40(11), 8287–8329. https://doi.org/10.1007/s00371-023-03237-7. Varney, N., Asari, V . K., & Graehling, Q.,

work page doi:10.1007/s00371-023-03237-7
[28]

Wang, Q., Chen, X., Zhang, Z., Meng, Y ., Gu, Y ., 2025a

https://doi.org/10.3390/rs17152731. Wang, Q., Chen, X., Zhang, Z., Meng, Y ., Gu, Y ., 2025a. Masking graph cross-convolution network for multispectral point cloud classification. IEEE Trans. Geosci. Remote Sens. PP, 1–1. https://doi.org/10.1109/TGRS.2025.3545783. Wang, X., Song, L., Feng, Y ., & Zhu, J., 2025b. S3F2Net: Spatial-Spectral-Structural Featur...

work page doi:10.3390/rs17152731 2025
[29]

A review of regional and global scale Land Use/Land Cover (LULC) mapping products generated from satellite remote sensing. ISPRS J. Photogramm. Remote Sens., 206, 311 –334. https://doi.org/10.1016/j.isprsjprs.2023.11.014. Wang, Y ., Sun, Y ., Liu, Z., Sarma, S. E., Bronstein, M. M., Solomon, J. M.,

work page doi:10.1016/j.isprsjprs.2023.11.014 2023
[30]

ACM Trans

Dynamic Graph CNN for learning on point clouds. ACM Trans. Graph. (TOG). Wu, L., Chen, Y ., Le, Y ., Qian, Y ., Zhang, D., Wang, L., 2024a. A high-precision fusion bathymetry of multi -channel waveform curvature for bathymetric LiDAR systems. Int. J. Appl. Earth Obs. Geoinf. 128, 103770. https://doi.org/10.1016/j.jag.2024.103770. Wu, X., Jiang, L., Wang, ...

work page doi:10.1016/j.jag.2024.103770 2024
[31]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

GrowSP: Unsupervised semantic segmentation of 3D point clouds. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 17619 –17629. https://doi.org/10.1109/CVPR52729.2023.01690. Zhao, H., Jiang, L., Jia, J., Torr, P. H. S., Koltun, V ., 2021a. Point Transformer. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR...

work page doi:10.1109/cvpr52729.2023.01690 2023
[32]

Zhou, W., Liu, K., Jin, W., Wang, Q., She, Y ., Y u, Y ., Ma, C., 2025b

https://doi.org/10.3390/rs17142411. Zhou, W., Liu, K., Jin, W., Wang, Q., She, Y ., Y u, Y ., Ma, C., 2025b. Advancements in deep learning for point cloud classification and segmentation: A comprehensive review. Comput. Graph. 130, 104238. https://doi.org/10.1016/j.cag.2025.104238. Zolanvari, I., Ruano, S., Rana, A., Cummins, A., Smolic, A., Da Silva, R.,...

work page doi:10.3390/rs17142411 2025

[1] [1]

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-2-2024

ESTATE: A Large Dataset of Under -Represented Urban Objects for 3D Point Cloud Classification. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-2-2024. https://doi.org/10.5194/isprs-archives-XLVIII-2-2024-25-2024. Bayrak, O. C., Remondino, F., & Uzar, M.,

work page doi:10.5194/isprs-archives-xlviii-2-2024-25-2024 2024

[2] [2]

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII -1/W3-2023

A new dataset and methodology for urban -scale 3D point cloud classification. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII -1/W3-2023. https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-1-2023. Carós, M., Just, A., Seguí, S., & Vitrià, J.,

work page doi:10.5194/isprs-archives-xlviii-1-w3-2023-1-2023 2023

[3] [3]

ISPRS Open Journal of Photogrammetry and Remote Sensing, 19, 100119

Enhancing point cloud semantic segmentation via scalable domain adaptation with LoRA-enabled PointNet++. ISPRS Open Journal of Photogrammetry and Remote Sensing, 19, 100119. https://doi.org/10.1016/j.ophoto.2026.100119. Chen, Z.,

work page doi:10.1016/j.ophoto.2026.100119 2026

[4] [4]

In: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Geospatial Week 2019, Enschede, NL

Hybrid orientation of airborne lidar point clouds and aerial images. In: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Geospatial Week 2019, Enschede, NL. Graham, B., Engelcke, M., & Maaten, L.,

work page 2019

[5] [5]

3D semantic segmentation with submanifold sparse convolutional networks. Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 9224–9232. https://doi.ieeecomputersociety.org/10.1109/CVPR.2018.00961. Han, X., Liu, C., Dong, Z., & Yang, B.,

work page doi:10.1109/cvpr.2018.00961 2018

[6] [6]

In: Proceedings of the IEEE/CVF Conference on Computer 25 Vision and Pattern Recognition, pp

RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 11105–11114. https://doi.org/10.1109/CVPR42600.2020.01112. Jing, Z., Guan, H., Zhao, P., Li, D., Y u, Y ., Zang, Y ., Wang, H., & Li, J.,

work page doi:10.1109/cvpr42600.2020.01112 2020

[7] [7]

IEEE Geosci

Toward Hyperspectral Lidar: Measurement of Spectral Backscatter Intensity With a Supercontinuum Laser Source. IEEE Geosci. Remote Sens. Lett. 4, 211–215. https://doi.org/10.1109/LGRS.2006.888848. Karila, K., Matikainen, L., Litkey, P., Hyyppä, J., & Puttonen, E.,

work page doi:10.1109/lgrs.2006.888848 2006

[8] [8]

International Journal of Remote Sensing, 40, 1 –19

The effect of seasonal variation on automated land cover mapping from multispectral airborne laser scanning data. International Journal of Remote Sensing, 40, 1 –19. https://doi.org/10.1080/01431161.2018.1528023. Kölle, M., Laupheimer, D., Schmohl, S., Haala, N., Rottensteiner, F., & Wegner, J.,

work page doi:10.1080/01431161.2018.1528023 2018

[9] [9]

ISPRS Open Journal of Photogrammetry and Remote Sensing, 1, 100001

The Hessigheim 3D (H3D) benchmark on semantic segmentation of high -resolution 3D point clouds and textured meshes from UA V LiDAR and Multi -View-Stereo. ISPRS Open Journal of Photogrammetry and Remote Sensing, 1, 100001. https://doi.org/10.1016/j.ophoto.2021.100001. Li, D., Shen, X., Guan, H., Yu, Y ., Wang, H., Zhang, G., & Li, J.,

work page doi:10.1016/j.ophoto.2021.100001 2021

[10] [10]

AGFP-Net: Attentive geometric feature pyramid network for land cover classification using airborne multispectral LiDAR data. Int. J. Appl. Earth Obs. Geoinf., 108, 102723. https://doi.org/10.1016/j.jag.2022.102723. Liao, Y ., Xie, J., & Geiger, A.,

work page doi:10.1016/j.jag.2022.102723 2022

[11] [11]

IEEE Transactions on Pattern Analysis and Machine Intelligence

KITTI 360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3D. IEEE Transactions on Pattern Analysis and Machine Intelligence. https://doi.org/10.1109/TPAMI.2022.3179507. Mandlburger, G., Otepka, J., Wagner, W., Karel, W., & Pfeifer, N.,

work page doi:10.1109/tpami.2022.3179507 2022

[12] [12]

https://www.researchgate.net/publication/237437116

Orientation and processing of airborne laser scanning data (OPALS) - Concept and first results of a comprehensive ALS software. https://www.researchgate.net/publication/237437116. Mikael Reichler, Josef Taher, Petri Manninen, Harri Kaartinen, Juha Hyyppä, & Antero Kukko,

work page arXiv

[13] [13]

ISPRS Open Journal of Photogrammetry and Remote Sensing, 12, 100061

Semantic segmentation of raw multispectral laser scanning data from urban environments with deep neural networks. ISPRS Open Journal of Photogrammetry and Remote Sensing, 12, 100061. https://doi.org/10.1016/j.ophoto.2024.100061. Morsy, S., Shaker, A., & El-Rabbany, A.,

work page doi:10.1016/j.ophoto.2024.100061 2024

[14] [14]

Nachtergaele, F

https://doi.org/10.3390/s17050958. Nachtergaele, F. O. F., & Licona-Manzur, C.,

work page doi:10.3390/s17050958

[15] [15]

National Land Survey of Finland, (n.d.)

The Land Degradation Assessment in Drylands (LADA) Project: Reflections on Indicators for Land Degradation Assessment. National Land Survey of Finland, (n.d.). Laser scanning data 5 p. National Land Survey of Finland. https://www.maanmittauslaitos.fi/en/maps-and-spatial-data/datasets-and-interfaces/product-descriptions/laser-scanning-data-5-p (accessed Oc...

work page 2025

[16] [16]

Land -cover classification of multispectral LiDAR data using CNN with optimized hyper -parameters. ISPRS J. Photogramm. Remote Sens., 166, 241 –254. https://doi.org/10.1016/j.isprsjprs.2020.05.022. Pan, S., Guan, H., Yu, Y ., Li, J., & Peng, D.,

work page doi:10.1016/j.isprsjprs.2020.05.022 2020

[17] [17]

Qi, C.R., Su, H., Mo, K., & Guibas, L.J., 2017a

https://doi.org/10.1117/12.849641. Qi, C.R., Su, H., Mo, K., & Guibas, L.J., 2017a. PointNet: Deep learning on point sets for 3D classification and segmentation . Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 652–660. https://doi.org/10.1109/CVPR.2017.16. Qi, C.R., Yi, L., Su, H., Guibas, L.J., 2017b: PointNet++: Deep hierarchical feature learni...

work page doi:10.1117/12.849641 2017

[18] [18]

Multi-Kernel Graph Structure Learning for Multispectral Point Cloud Classification. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., PP, 1–15. https://doi.org/10.1109/JSTARS.2024.3368472. Riveiro, B., Lindenbergh, R.,

work page doi:10.1109/jstars.2024.3368472 2024

[19] [19]

ISPRS Journal of Photogrammetry and Remote Sensing, 236, 569–605

Benchmarking individual tree segmentation using multispectral airborne laser scanning data: The FGI -EMIT dataset. ISPRS Journal of Photogrammetry and Remote Sensing, 236, 569–605. https://doi.org/10.1016/j.isprsjprs.2026.04.021. Ruoppa, L., Oinonen, O., Taher, J., Lehtomäki, M., Takhtkeshha, N., Kukko, A., Kaartinen, H., Hyyppä, J.,

work page doi:10.1016/j.isprsjprs.2026.04.021 2026

[20] [20]

Unsupervised deep learning for semantic segmentation of multispectral LiDAR forest point clouds. ISPRS J. Photogramm. Remote Sens. 228, 694–722. https://doi.org/10.1016/j.isprsjprs.2025.07.038. Saurabh Prasad, Bertrand Le Saux, Naoto Yokoya, & Ronny Hansch,

work page doi:10.1016/j.isprsjprs.2025.07.038 2025

[21] [21]

IEEE Dataport

2018 IEEE GRSS Data Fusion Challenge – Fusion of Multispectral LiDAR and Hyperspectral Data. IEEE Dataport. https://doi.org/10.21227/jnh9-nz89. Serna, A., Marcotegui, B., Goulette, F., & Deschaud, J. -E.,

work page doi:10.21227/jnh9-nz89 2018

[22] [22]

ICPRAM 2014, 1–4

Paris-rue-Madame database: a 3D mobile laser scanner dataset for benchmarking urban detection, segmentation and classification methods. ICPRAM 2014, 1–4. Taher, J., Hakala, T., Jaakkola, A., Hyyti, H., Kukko, A., Manninen, P., Maanpää, J., Hyyppä, J.,

work page 2014

[23] [23]

Taher, J., Hyyppä, E., Hyyppä, M., Salolahti, K., Y u, X., Matikainen, L., Kukko, A., Lehtomäki, M., Kaartinen, H., Thurachen, S., et al.,

https://doi.org/10.3390/s22155759. Taher, J., Hyyppä, E., Hyyppä, M., Salolahti, K., Y u, X., Matikainen, L., Kukko, A., Lehtomäki, M., Kaartinen, H., Thurachen, S., et al.,

work page doi:10.3390/s22155759

[24] [24]

Multispectral airborne laser scanning for tree species classification: A benchmark of machine learning and deep learning algorithms. ISPRS J. Photogramm. Remote Sens. 233, 278–309. https://doi.org/10.1016/j.isprsjprs.2026.01.031. Takhtkeshha, N., Bayrak, O.C., Mandlburger, G., Remondino, F., Kukko, A., Hyyppä, J.,2024a. Automatic annotation of 3D multispe...

work page doi:10.1016/j.isprsjprs.2026.01.031 2026

[25] [25]

Takhtkeshha, N., Mazzacca, G., Remondino, F., Hyyppä, J., & Mandlburger, G., 2025b

https://doi.org/10.3390/s24051669. Takhtkeshha, N., Mazzacca, G., Remondino, F., Hyyppä, J., & Mandlburger, G., 2025b. Multispectral LiDAR data for extracting tree points in urban and suburban areas. https://doi.org/10.48550/arXiv.2508.19881. Thomas, H., Qi, C. R., Deschaud, J. -E., Marcotegui, B., Goulette, F., Guibas, L.,

work page doi:10.3390/s24051669

[26] [26]

PoseNet: A convolutional network for real-time 6-dof camera relocalization,

KPConv: Flexible and deformable convolution for point clouds. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, South Korea, pp. 6410–6419. https://doi.org/10.1109/ICCV .2019.00651. Thomas, H., Tsai, Y .-H.H., Barfoot, T.D., Zhang, J.,

work page doi:10.1109/iccv 2019

[27] [27]

The Visual Computer, 40(11), 8287–8329

Deep learning based computer vision under the prism of 3D point clouds: a systematic review. The Visual Computer, 40(11), 8287–8329. https://doi.org/10.1007/s00371-023-03237-7. Varney, N., Asari, V . K., & Graehling, Q.,

work page doi:10.1007/s00371-023-03237-7

[28] [28]

Wang, Q., Chen, X., Zhang, Z., Meng, Y ., Gu, Y ., 2025a

https://doi.org/10.3390/rs17152731. Wang, Q., Chen, X., Zhang, Z., Meng, Y ., Gu, Y ., 2025a. Masking graph cross-convolution network for multispectral point cloud classification. IEEE Trans. Geosci. Remote Sens. PP, 1–1. https://doi.org/10.1109/TGRS.2025.3545783. Wang, X., Song, L., Feng, Y ., & Zhu, J., 2025b. S3F2Net: Spatial-Spectral-Structural Featur...

work page doi:10.3390/rs17152731 2025

[29] [29]

A review of regional and global scale Land Use/Land Cover (LULC) mapping products generated from satellite remote sensing. ISPRS J. Photogramm. Remote Sens., 206, 311 –334. https://doi.org/10.1016/j.isprsjprs.2023.11.014. Wang, Y ., Sun, Y ., Liu, Z., Sarma, S. E., Bronstein, M. M., Solomon, J. M.,

work page doi:10.1016/j.isprsjprs.2023.11.014 2023

[30] [30]

ACM Trans

Dynamic Graph CNN for learning on point clouds. ACM Trans. Graph. (TOG). Wu, L., Chen, Y ., Le, Y ., Qian, Y ., Zhang, D., Wang, L., 2024a. A high-precision fusion bathymetry of multi -channel waveform curvature for bathymetric LiDAR systems. Int. J. Appl. Earth Obs. Geoinf. 128, 103770. https://doi.org/10.1016/j.jag.2024.103770. Wu, X., Jiang, L., Wang, ...

work page doi:10.1016/j.jag.2024.103770 2024

[31] [31]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

GrowSP: Unsupervised semantic segmentation of 3D point clouds. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 17619 –17629. https://doi.org/10.1109/CVPR52729.2023.01690. Zhao, H., Jiang, L., Jia, J., Torr, P. H. S., Koltun, V ., 2021a. Point Transformer. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR...

work page doi:10.1109/cvpr52729.2023.01690 2023

[32] [32]

Zhou, W., Liu, K., Jin, W., Wang, Q., She, Y ., Y u, Y ., Ma, C., 2025b

https://doi.org/10.3390/rs17142411. Zhou, W., Liu, K., Jin, W., Wang, Q., She, Y ., Y u, Y ., Ma, C., 2025b. Advancements in deep learning for point cloud classification and segmentation: A comprehensive review. Comput. Graph. 130, 104238. https://doi.org/10.1016/j.cag.2025.104238. Zolanvari, I., Ruano, S., Rana, A., Cummins, A., Smolic, A., Da Silva, R.,...

work page doi:10.3390/rs17142411 2025