Datasets:
MatchGeo Dataset — Version 1.1
1. Dataset Overview
| Attribute | Details |
|---|---|
| Title | MatchGeo: Multi-region Digital Elevation Model Dataset for Local Feature Matching |
| Version | 1.1 |
| Release Date | 2026-05-11 |
| Authors / Creators | Correa, S. P. L. P.; Pazini Pedro, D. F.; Oliveira, H. N.; Belton, D.; Ivánová, I.; Santos, A. de Paula |
| Contact | sabrina.correa@ufv.br |
| Persistent Identifier (DOI) | 10.5281/zenodo.21229785 |
| Keywords | digital elevation model, DEM, DSM, local feature matching, geospatial dataset, urban terrain, cross-domain matching, computer vision, GeoTIFF, LiDAR, photogrammetry, Structure-from-Motion, satellite stereophotogrammetry |
| Related Publication | In preparation |
| Repository | Zenodo · Hugging Face |
2. Description
MatchGeo is a curated, multi-region Digital Elevation Model (DEM) dataset designed for training and benchmarking local feature matching algorithms in urban and natural terrain analysis. It aggregates high-resolution elevation data from 13 distinct environments across 6 continents to support research in cross-domain local feature detection and matching under varying acquisition methods, climates, terrain types, and urban morphologies.
The dataset provides standardized 256x256 pixel GeoTIFF patches with handcrafted ground truth annotations for two cities, enabling systematic evaluation of both intra-region and inter-region model generalization.
Geographic & Temporal Scope
| Region | Country | Temporal Coverage | Spatial Coverage | Terrain Type |
|---|---|---|---|---|
| Antarctic Peninsula (ATA_MV) | Antarctica | 2009–2024 | Peninsula | Polar, ice |
| São Paulo (BRA_SP) | Brazil | 2020 | Municipal (1,544 km²) | Tropical, urban |
| Wutai Shan (CHN_WS) | China | 2021 | Regional | Mountainous |
| El Hierro (ESP_EH) | Canary Islands (Spain) | 2022–2025 | Regional | Volcanic, coastal |
| Lahti Lake (FIN_LM) | Finland | 2020–2026 | Regional | Temperate, country |
| Bonn (GER_BN) | Germany | 2016–2018 | City limits | Temperate, country |
| Sinabung Volcano (IDN_SV) | Indonesia | 2018 | 16.07 km² | Volcanic, tropical |
| Almaty City (KAZ_AC) | Kazakhstan | 2017 | 304.28 km² | Semi-arid, urban |
| Wadi Al-Akhdar (KSA_WA) | Saudi Arabia | 2016 | 1,200 km² | Desert, graben |
| Hebron Fault (NAM_HF) | Namibia | 2017 | Regional | Arid, fault zone |
| Kapiti Coast (NZL_KP) | New Zealand | 2010–2025 | Regional | Coastal, temperate, country |
| Tarlac (PHL_TA) | Philippines | 2014–2017 | Regional | Tropical, flat |
| Grand Canyon (USA_GC) | United States | 2020–2026 | Regional | Desert, canyon |
Purpose & Use Cases
- Training and benchmarking local feature detectors and descriptors on geospatial elevation data
- Cross-domain generalization studies (LiDAR vs. photogrammetry vs. satellite, temperate vs. tropical vs. desert)
- Urban terrain analysis and change detection
- Evaluation of matching robustness across sensor modalities and terrain types
- Benchmarking for computer vision models on non-RGB data
3. Content & Schema
3.1 Files Included
| Filename / Folder | Description | Approx. Size |
|---|---|---|
data/ATA_MV/ |
Antarctic Peninsula tiles + metadata | 138.3 MB |
data/BRA_SP/ |
São Paulo tiles + annotations + metadata | 105.34 MB |
data/CHN_WS/ |
Wutai Shan tiles + metadata | 206.08 MB |
data/ESP_EH/ |
El Hierro tiles + metadata | 422.49 MB |
data/FIN_LM/ |
Lahti Lake tiles + metadata | 45.97 MB |
data/GER_BN/ |
Bonn tiles + annotations + metadata | 333.24 MB |
data/IDN_SV/ |
Sinabung Volcano tiles + metadata | 34.98 MB |
data/KAZ_AC/ |
Almaty City tiles + metadata | 128.00 MB |
data/KSA_WA/ |
Wadi Al-Akhdar tiles + metadata | 556.85 MB |
data/NAM_HF/ |
Hebron Fault tiles + metadata | 115.24 MB |
data/NZL_KP/ |
Kapiti Coast tiles + metadata | 362.23 MB |
data/PHL_TA/ |
Tarlac tiles + metadata | 64.10 MB |
data/USA_GC/ |
Grand Canyon tiles + metadata | 138.27 MB |
splits/ |
Train / validation / test split manifests (CSV) | 1.6 MB |
manifest.json |
Central machine-readable catalog (JSON-LD) | 7.2 KB |
README.md |
High-level project documentation | 12.5 KB |
DATASET_DESCRIPTION.md |
This file | 126.0 KB |
LICENSE |
Full CC BY 4.0 legal text | 14.2 KB |
3.2 Per-region File Structure
Each region folder follows this structure:
<REGION_ID>/
├── <REGION_ID>.tif # Merged DEM (BigTIFF, tiled, DEFLATE)
├── <REGION_ID>_extent.geojson # Bounding polygon of full coverage
├── <REGION_ID>_tiles.geojson # Tile index (grid of all 256x256 patches)
├── metadata/
│ └── <REGION_ID>_metadata.json # ISO 19115-2 + OGC 23-008r3 metadata
├── annotations/ # Only for labelled cities (BRA_SP, GER_BN)
│ └── <REGION_ID>_###_###.json # Keypoint annotations per tile
└── tiles/
└── <REGION_ID>_###_###.tif # 256x256 pixel patches
3.3 Data Dictionary
| Variable / Field | Data Type | Description | Units / Format | Nullable | Example |
|---|---|---|---|---|---|
tile_id |
string | Unique identifier for each 256x256 patch | — | No | GER_BN_001_010 |
region |
string | Source region code | — | No | GER_BN |
geometry |
GeoTIFF raster | Elevation patch | 256x256 px, float32 | No | — |
crs |
string | Coordinate reference system | EPSG code | No | EPSG:25832 |
resolution |
float | Ground sample distance | meters | No | 1.0 |
acquisition_method |
string | Data acquisition technique | — | No | airborne_lidar |
acquisition_year |
integer | Year of data capture | year | No | 2017 |
elevation_min |
float | Minimum elevation in patch | meters | Yes | 42.3 |
elevation_max |
float | Maximum elevation in patch | meters | Yes | 156.8 |
elevation_mean |
float | Mean elevation in patch | meters | Yes | 98.4 |
elevation_std |
float | Standard deviation of elevation | meters | Yes | 12.1 |
annotated |
boolean | Whether patch has handcrafted ground truth | — | No | true |
n_annotations |
integer | Number of verified point annotations in patch | count | Yes | 127 |
split |
string | Dataset split assignment | — | No | train |
difficulty |
string | Difficulty tag for stratification | — | Yes | easy |
4. Methodology & Provenance
4.1 Data Sources
All raw data were obtained from open municipal, national, or research portals and are redistributed under terms compatible with CC BY 4.0.
| City | Source Dataset | Provider | Method | Resolution | Year | CRS | License / Terms |
|---|---|---|---|---|---|---|---|
| ATA_MV | REMA | Polar Geospatial Center / University of Minnesota | Satellite stereophotogrammetry | 1.0 m | 2009–2024 | EPSG:3031 | CC BY 4.0 |
| BRA_SP | GeoSampa MDS 2020 | Prefeitura de São Paulo (PMSP) / SMUL / GEOINFO | Airborne LiDAR (Optech ORION H300) | 0.5 m | 2020 | EPSG:31983 | Open municipal data |
| CHN_WS | Wutai Shan / Yingwang Shan 2021 | Zhou, C. / OpenTopography | UAV SfM (DJI Phantom 4) | 1.0 m | 2021 | EPSG:32649 | OpenTopography terms |
| ESP_EH | PNOA-LiDAR 3ª Cobertura | CNIG / Instituto Geográfico Nacional | Airborne LiDAR | 0.5 m | 2022–2025 | EPSG:3040 | CC BY 4.0 |
| FIN_LM | Elevation Model 2 m | Maanmittauslaitos (NLS Finland) | Airborne LiDAR + photogrammetry | 2.0 m | 2020–2026 | EPSG:3067 | CC BY 4.0 |
| GER_BN | Digitales Oberflächenmodell (DOM) | Geobasis NRW / Bezirksregierung Köln | Airborne LiDAR | 1.0 m | 2016–2018 | EPSG:25832 | Open data |
| IDN_SV | Sinabung Volcano 2018 | Carr, B. / OpenTopography | UAS SfM (DJI Matrice 210) | 0.87 m | 2018 | EPSG:32647 | OpenTopography terms |
| KAZ_AC | Almaty City 2017 | Amey et al. / OpenTopography | Pleiades Tristereo | 1.0 m | 2017 | EPSG:32643 | OpenTopography terms |
| KSA_WA | Wadi-al-Akhdar 2016 | Matthieu et al. / OpenTopography | SPOT 6 Stereo | 1.6 m | 2016 | EPSG:32637 | OpenTopography terms |
| NAM_HF | Hebron Fault 2017 | Salomon et al. / UCT / OpenTopography | WorldView-3 Stereo | 0.53 m | 2017 | EPSG:32733 | OpenTopography terms |
| NZL_KP | NZ LiDAR 1m DEM | LINZ | Airborne LiDAR | 1.0 m | 2010–2025 | EPSG:2193 | CC BY 4.0 |
| PHL_TA | LiPAD | UP Diliman TCAGP / DREAM / DOST | Airborne LiDAR | 1.0 m | 2014–2017 | EPSG:32651 | Open data |
| USA_GC | USGS 3DEP DEM | U.S. Geological Survey | LiDAR | 0.5 m | 2020–2026 | EPSG:6341 | Public Domain |
4.2 Processing Pipeline
All cities were processed through a standardized PDAL pipeline with region-specific adaptations:
- Acquisition & Download — Raw data retrieved from source portals in native CRS and resolution.
- Preprocessing (region-specific):
- BRA_SP: Ground classification via
filters.smf(scalar=1.25, slope=0.15, threshold=0.5, window=16.0) - CHN_WS: Noise removal (
filters.rangeClassification![7:7]) + statistical outlier removal (mean_k=6, multiplier=2.0) - IDN_SV, KSA_WA, NAM_HF: Statistical outlier removal (mean_k=6, multiplier=2.0)
- Others: Direct rasterization
- BRA_SP: Ground classification via
- Rasterization — PDAL
writers.gdalwith:output_type:max(DSM: highest point per cell)data_type:float32nodata:-9999gdalopts:COMPRESS=DEFLATE|TILED=YES|BIGTIFF=YESresolution: City-specific (see table above)
- Patch Extraction — Non-overlapping 256x256 pixel patches extracted using a regular grid (no resampling).
- Quality Control — Outlier removal based on elevation range checks, CRS validation, and visual inspection.
- Annotation (BRA_SP, GER_BN) — Handcrafted point-level annotations in normalized coordinates.
- Stratified Splitting — 80% train / 10% validation / 10% test (planned; to be stratified by region and difficulty).
- Packaging & Metadata — Per-region ISO 19115-2 metadata; central JSON-LD manifest; archived on Zenodo with DOI.
4.3 Software & Tools
- Python 3.10.20
- PDAL 2.6.0 (point cloud processing)
- GDAL / rasterio (geospatial I/O and reprojection)
- NumPy / pandas (tabular metadata management)
- Custom annotation and patch-generation scripts
5. Access & Licensing
5.1 Availability
| Repository | URL | Access Type |
|---|---|---|
| Zenodo (Primary) | https://doi.org/10.5281/zenodo.21229785 | Open, DOI-backed, permanent |
| Hugging Face Datasets | https://huggingface.co/datasets/paeslemesa/matchgeo | Open, streaming loader available |
- Total Size: ~12.5 GB
- File Formats: Cloud Optimized GeoTIFF - COG (
.tif), GeoJSON (.geojson), metadata JSON, split manifests CSV - Access Type: Open access, no registration required
5.2 License
This dataset is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
You are free to:
- Share — copy and redistribute the material in any medium or format
- Adapt — remix, transform, and build upon the material for any purpose, even commercially
Under the following terms:
- Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made.
The full legal text is provided in the LICENSE file and at https://creativecommons.org/licenses/by/4.0/legalcode.
5.3 Terms of Use & Source Attribution
When using this dataset, your publication or product must include:
- Dataset citation (see Section 8: Citation)
- Original source acknowledgments (per region):
- ATA_MV: Data derived from REMA © Polar Geospatial Center / University of Minnesota
- BRA_SP: Data derived from GeoSampa © Prefeitura de São Paulo
- CHN_WS: Data derived from OpenTopography dataset by Zhou, C. (DOI: 10.5069/G98C9TGT)
- ESP_EH: Data derived from PNOA-LiDAR © CNIG / Instituto Geográfico Nacional
- FIN_LM: Data derived from Maanmittauslaitos © National Land Survey of Finland
- GER_BN: Data derived from Geobasis NRW © Bezirksregierung Köln
- IDN_SV: Data derived from OpenTopography dataset by Carr, B. (DOI: 10.5069/G8988568)
- KAZ_AC: Data derived from OpenTopography dataset by Amey et al. (DOI: 10.5069/G9H41PMP)
- KSA_WA: Data derived from OpenTopography dataset by Matthieu et al. (DOI: 10.5069/G9V40SDZ)
- NAM_HF: Data derived from OpenTopography dataset by Salomon et al. (DOI: 10.5069/G9W957BC)
- NZL_KP: Data derived from LINZ © Land Information New Zealand
- PHL_TA: Data derived from LiPAD © UP Diliman TCAGP / DREAM Program
- USA_GC: Data derived from USGS 3DEP © U.S. Geological Survey
6. Interoperability & Technical Details
6.1 Standards & Formats
| Aspect | Standard / Value |
|---|---|
| Raster format | Cloud Optimized GeoTIFF (BigTIFF variant, OGC 23-008r3 compliant) |
| Internal tiling | 256 × 256 pixels |
| Pixel depth | Float32 |
| Patch dimensions | 256 × 256 pixels |
| Coordinate systems | Region-specific UTM (see Section 4.1) |
| Metadata standard | ISO 19115-2 + OGC 23-008r3 |
| Encoding | UTF-8 for all text and tabular files |
| Compression | DEFLATE |
| NoData value | -9999 |
6.2 OGC GeoTIFF Compliance (23-008r3)
All GeoTIFFs comply with OGC 23-008r3 and contain the following required keys:
| Key | Status | Source |
|---|---|---|
| GTModelTypeGeoKey | ✅ | PDAL override_srs |
| GTRasterTypeGeoKey | ✅ | GDAL default (PixelIsArea) |
| ProjectedCSTypeGeoKey | ✅ | EPSG code embedded |
| GeogGeodeticDatumGeoKey | ✅ | Derived from EPSG |
| GeogAngularUnitsGeoKey | ✅ | Degree (default) |
| ProjLinearUnitsGeoKey | ✅ | Meter (default) |
| PixelScale | ✅ | GDAL writer |
| TiePoint | ✅ | Upper-left corner coordinate |
| BigTIFF | ✅ | BIGTIFF=YES |
| Tiled | ✅ | TILED=YES |
| Compression | ✅ | COMPRESS=DEFLATE |
| NoData | ✅ | nodata=-9999 |
Note on vertical CRS: The GeoTIFFs encode 2D projected CRS (EPSG codes) with elevation as pixel values. A vertical CRS is not explicitly embedded in the GeoTIFF GeoKeys; vertical datum information is documented in per-region
metadata.jsonfiles.
6.3 Controlled Vocabularies
- GCMD Keywords:
EARTH SCIENCE > LAND SURFACE > TOPOGRAPHY > TERRAIN ELEVATION > DIGITAL ELEVATION/TERRAIN MODEL (DEM) - ISO Topic Category:
elevation - INSPIRE Theme:
Elevation
6.4 Data Splits
| Split | Proportion | Stratification |
|---|---|---|
| Train | 80% | By region and difficulty |
| Validation | 10% | By region and difficulty |
| Test | 10% | By region and difficulty |
Explicit intra-region and inter-region test subsets are planned for future releases.
7. Quality Assurance & Known Limitations
7.1 Validation
- Bonn annotations were verified by multiple annotators.
- Elevation ranges were cross-checked against known region topographies.
- CRS consistency was validated using GDALinfo and automated assertions.
7.2 Known Biases & Limitations
| Issue | Description | Mitigation / Status |
|---|---|---|
| Geographic bias | Dense annotation only in GER_BN and BRA_SP; others unlabelled | v1.1 separates labelled/unlabelled; future releases will add annotations |
| Temporal mismatch | Data spans 2009–2026 across cities | Documented; users should account for temporal drift |
| Sensor heterogeneity | LiDAR, photogrammetry, SfM, satellite stereo — different noise characteristics | Explicitly treated as cross-domain challenge |
| Resolution heterogeneity | Native resolutions range from 0.5 m to 2.0 m | |
| Missing data | Water bodies, ocean excluded; NoData=-9999 | Documented in per-tile metadata |
| ATA_MV uncertainty | Antarctica data from REMA satellite stereophotogrammetry | Marked for verification |
| CHN_WS CRS discrepancy | BibTeX says EPSG:32649; user previously mentioned EPSG:32648 | To be verified from actual file |
| Vertical CRS | No explicit vertical GeoKey in GeoTIFFs; elevations are source-dependent (orthometric or ellipsoidal) | Documented in per-region metadata.json |
8. Citation
If you use this dataset in your research, please cite:
@dataset{correa_2026_matchgeo,
author = {Correa, S. P. L. P. and Pazini Pedro, D. F. and Oliveira, H. N. and Belton, D. and Ivánová, I. and Santos, A. de Paula},
title = {{MatchGeo: Digital Elevation Model Dataset for Local Feature Matching}},
year = 2026,
publisher = {Zenodo},
version = {1.1},
doi = {10.5281/zenodo.21229785},
url = {https://doi.org/10.5281/zenodo.21229785},
note = {Contains data derived from REMA, GeoSampa, OpenTopography, CNIG, Maanmittauslaitos, Geobasis NRW, LINZ, LiPAD, and USGS 3DEP}
}
Plain text citation:
Correa, S. P. L. P., Pazini Pedro, D. F., Oliveira, H. N., Belton, D., Ivánová, I., & Santos, A. de Paula. (2026). MatchGeo: Digital Elevation Model Dataset for Local Feature Matching (Version 1.1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.21229785
Source Dataset Citations
When using specific cities, also cite the original sources:
% BRA_SP
@misc{datasetgeosampa,
author = {{Prefeitura do Munic{'i}pio de S{\~a}o Paulo (PMSP)}},
title = {{Nuvem de Pontos {MDS} 2020}},
year = {2020},
publisher = {GeoSampa},
url = {https://geosampa.prefeitura.sp.gov.br/}
}
% CHN_WS
@misc{datasetchinayingwang2021,
author = {Zhou, C.},
title = {High Resolution Topography of {Wutai Shan} and {Yingwang Shan}, China, 2021},
year = {2023},
publisher = {OpenTopography},
doi = {10.5069/G98C9TGT}
}
% ESP_EH
@misc{datasetspain,
author = {{Centro Nacional de Informaci{'o}n Geogr{'a}fica (CNIG)}},
title = {{PNOA-LiDAR --- 3{\textordfeminine} Cobertura (2022--2025)}},
year = {2022--2025},
url = {https://centrodedescargas.cnig.es/CentroDescargas/lidar-tercera-cobertura}
}
% FIN_LM
@misc{datasetfinland,
author = {{Maanmittauslaitos --- National Land Survey of Finland}},
title = {{Korkeusmalli 2 m / Elevation Model 2 m}},
year = {2026},
url = {https://www.maanmittauslaitos.fi/en/maps-and-spatial-data/datasets-and-interfaces/product-descriptions/elevation-model-2-m}
}
% GER_BN
@misc{datasetgermany,
author = {{Geobasis NRW --- Bezirksregierung K{\"o}ln}},
title = {{Digitales Oberfl{\"a}chenmodell ({DOM})}},
year = {2024},
url = {https://www.bezreg-koeln.nrw.de/geobasis-nrw/produkte-und-dienste/hoehenmodelle/digitale-oberflaechenmodelle/digitales}
}
% IDN_SV
@misc{datasetindonesiasinabung2018,
author = {Carr, B.},
title = {{Sinabung Volcano} (Indonesia), {June 20, 2018}},
year = {2021},
publisher = {OpenTopography},
doi = {10.5069/G8988568}
}
% KAZ_AC
@misc{datasetkazakhstanalmaty2017,
author = {Amey, R. and Watson, C. S. and Elliott, J. and Walker, R.},
title = {{Almaty City, Kazakhstan, 2017, Derived from {Pleiades} Tristereo Imagery}},
year = {2021},
publisher = {OpenTopography},
doi = {10.5069/G9H41PMP}
}
% KSA_WA
@misc{dataset_saudiarabia_wadi_al_akhdar_2016,
author = {Matthieu, R. and Moulin, A. and J{'o}nsson, S.},
title = {{Digital Surface Model of {Wadi-al-Akhdar} Graben, Saudi Arabia, 2016}},
year = {2024},
publisher = {OpenTopography},
doi = {10.5069/G9V40SDZ}
}
% NAM_HF
@misc{datasetnamibiahebron_2017,
author = {Salomon, G. and Smit, J. and Muir, R. and Stevens, V. and Sloan, R. A.},
title = {{Hebron Fault, Namibia 2017 {WorldView-3} Stereophotogrammetric {DEM}}},
year = {2021},
publisher = {OpenTopography},
doi = {10.5069/G9W957BC}
}
% NZL_KP
@misc{datasetnewzealand,
author = {{Toit{\=u} Te Whenua Land Information New Zealand ({LINZ})}},
title = {New Zealand {LiDAR} 1m {DEM}},
year = {2025},
url = {https://data.linz.govt.nz/layer/121859-new-zealand-lidar-1m-dem/services/csw/}
}
% PHL_TA
@misc{datasetphilippines,
author = {{Disaster Risk and Exposure Assessment for Mitigation (DREAM) Program}},
title = {{LiPAD --- {LiDAR} Portal for Archiving and Distribution}},
year = {2016},
url = {https://lipad.dream.upd.edu.ph/}
}
% USA_GC
@misc{datasetusa,
author = {{U.S. Geological Survey}},
title = {{3D Elevation Program (3DEP) Digital Elevation Models}},
year = {2026},
url = {https://apps.nationalmap.gov/downloader/}
}
9. Version History
| Version | Date | Changes | Author |
|---|---|---|---|
| 1.0 | 2026-03-30 | Initial release; GER_BN 20,000+ handcrafted annotations | Correa et al. |
| 1.1 | 2026-05-11 | Expanded to 13 cities; reorganized into labelled/unlabelled; added per-region metadata; synchronized with FAIR + OGC 23-008r3 | Correa et al. |
10. Acknowledgments & Funding
- Data providers: European Commission / ESA (Copernicus), Prefeitura de São Paulo (GeoSampa), OpenTopography, CNIG (Spain), Maanmittauslaitos (Finland), Geobasis NRW (Germany), LINZ (New Zealand), UP Diliman TCAGP / DREAM (Philippines), USGS (United States)
- Imagery providers: CNES / Airbus DS (Pleiades), Maxar (WorldView-3), SPOT Image (SPOT 6)
- Institutional support: Universidade Federal de Viçosa (UFV)
11. Contact & Support
- General inquiries: sabrina.correa@ufv.br
- Dataset maintainer: Sabrina Correa, Universidade Federal de Viçosa
- Issues / bug reports: GitHub Issues
- Discussions: GitHub Discussions
Last Updated: 2026-07-09