ConRE-OSM

Continuity-based Road Extraction from OSM

OSM Road Surface Reconstruction — Case Study: Delft

Reconstruct lane-level road and intersection polygons from OpenStreetMap (OSM) centerline data for Delft, Netherlands.

This repository contains the source code of the ConRE-OSM tool, developed as part of a thesis study using road centerline data from OpenStreetMap for the city of Delft, Netherlands. ConRE-OSM reconstructs topologically continuous, lane-level road surface polygons and intersection areas directly from OSM way geometry, without requiring additional surveyed or manually prepared inputs. The pipeline derives lane offsets, corrects geometric connectivity at forks and bi-directional junctions, and generates semantically labelled road and intersection polygons that preserve the continuity and topology of the underlying street network.

Citation

This repository accompanies the MSc thesis:

Rao, C. (2024). Reconstructing a high-detailed 2D areal representation of road network based on OSM data [Master's thesis, Delft University of Technology]. TU Delft Repository. https://resolver.tudelft.nl/uuid:0bae72e4-ff16-4921-8f6e-d92fa4a377d7

If you use this work ConRE-OSM, please cite the following:

Thesis

@mastersthesis{rao2024roadreconstructing,
  author  = {Rao, Chengzhi},
  title   = {Reconstructing a high-detailed 2D areal representation of road network based on OSM data},
  school  = {Delft University of Technology},
  year    = {2024},
  type    = {Master's thesis},
  note    = {Faculty of Architecture and the Built Environment, Programme: Geomatics},
  url     = {https://resolver.tudelft.nl/uuid:0bae72e4-ff16-4921-8f6e-d92fa4a377d7}
}

Tool

@software{conre_osm_2026,
  author  = {[Chengzhi Rao]},
  title   = {{ConRE-OSM}: Continuity-based Road Extraction from OSM},
  year    = {2026},
  url     = {[repository URL]},
  license = {GPL-3.0-or-later}
}

References

The following works informed the design of ConRE-OSM:

Methods and tools used

osmnx — Boeing, G. (2017). OSMnx: New Methods for Acquiring, Constructing, Analyzing, and Visualizing Complex Street Networks. Computers, Environment and Urban Systems, 65, 126–139. https://doi.org/10.1016/j.compenvurbsys.2017.05.004
momepy — Fleischmann, M. (2019). momepy: Urban Morphology Measuring Toolkit. Journal of Open Source Software, 4(43), 1807. https://doi.org/10.21105/joss.01807
GeoPandas — Jordahl, K. et al. (2014). GeoPandas: Python tools for geographic data. https://github.com/geopandas/geopandas
Shapely — Gillies, S. et al. (2007). Shapely: manipulation and analysis of geometric objects. https://github.com/shapely/shapely

Related projects

COINS — Tripathy, P., et al. (2020). An open-source tool to extract natural continuity and hierarchy of urban street networks. Environment and Planning B. https://doi.org/10.1177/2399808320967680
StreetGen — Martin, S., et al. (2015). StreetGen: In-base, city-scale procedural generation of streets. Proceedings of the Eurographics Workshop on Urban Data Modelling and Visualisation.
A/B Street — https://github.com/a-b-street/abstreet
OpenStreetMap — OpenStreetMap contributors (2017). Planet dump retrieved from https://planet.osm.org, https://www.openstreetmap.org

Final outputs

File	Description	Rows
`data/output/road_intersections.geojson`	Intersection polygons	108
`data/output/road_segments.geojson`	Road polygons with intersections removed	329

All files use EPSG:28992 (RD New, metres).

Setup

pip install -r requirements.txt

Python 3.10+. Place your OSM export at:

data/input/delft_osm_ways.geojson

This is the only external input required. It should be an OSM ways export in WGS84 (EPSG:4326) covering the study area. Filtering to the required highway types and reprojection to EPSG:28992 happen automatically inside the pipeline.

Run

python run.py               # run all 5 stages
python run.py --stages 1 2  # run only stages 1 and 2
python run.py --from 3      # run stages 3 through 5
python run.py --no-mid      # skip saving mid-results (faster)

All output paths are defined at the top of run.py. Change them there to relocate data.

Pipeline overview

data/input/delft_osm_ways.geojson          ──► Stage 1, Stage 4, Stage 5

Stage 1 ──► data/output/stage/s1_road_attributes.geojson  ──► Stage 2, 3, 4
Stage 2 ──► data/output/stage/s2_lane_centerlines.geojson ──► Stage 3
         ──► data/output/stage/s2_lane_polygons.geojson   ──► Stage 3
         ──► data/output/stage/s2_lane_boundaries.geojson ──► Stage 3
         ──► data/output/stage/s2_road_dict.json          ──► Stage 3
Stage 3 ──► data/output/stage/s3_lane_segments.geojson    ──► Stage 4, Stage 5
Stage 4 ──► data/output/stage/s4_road_polygons.geojson    ──► Stage 5
         ──► data/output/stage/s4_network_nodes.geojson   ──► Stage 5
Stage 5 ──► data/output/road_intersections.geojson        (FINAL)
         ──► data/output/road_segments.geojson            (FINAL)

Data flow — detailed

Stage 1 — Extract, Classify and Label Road Centerlines

Input: data/input/delft_osm_ways.geojson

Stage result → stages 2, 3, 4:

File	Rows	Role
`data/output/stage/s1_road_attributes.geojson`	554	Road attribute table: lane counts, turn:lanes, bearings, group_index, oneway, fork_type per road segment

Mid results (data/output/mid/):

File	After which step
`s1_cycle_extract.geojson`	Cycleway extraction from parent-road attributes
`s1_backward_split.geojson`	Bi-directional roads split into `_forward`/`_backward`
`s1_turn_normalized.geojson`	`turn:lanes` normalised to pipe-separated strings
`s1_grouped.geojson`	Initial grouping (superseded by stage result)

Stage 2 — Offset Lane Centerlines

Input: data/output/stage/s1_road_attributes.geojson

Stage results → stage 3:

File	Description
`data/output/stage/s2_lane_centerlines.geojson`	Offset lane centerlines (`id_raw`, `lane_type`, `lane_order`, geometry)
`data/output/stage/s2_lane_polygons.geojson`	Buffered lane polygons, same structure
`data/output/stage/s2_lane_boundaries.geojson`	Two outermost boundary lines per road
`data/output/stage/s2_road_dict.json`	Per-road lane data (393 entries): ids, types, turns, offsets, widths, groups

Mid results: none.

Stage 3 — Modify Lane Geometry for Topology

Inputs: s1_road_attributes.geojson + all s2_* files (stage 2)

Stage result → stages 4, 5:

File	Rows	Role
`data/output/stage/s3_lane_segments.geojson`	659	Per-lane metadata with corrected geometry: `id_raw`, `id_lane`, `type`, `highway`, `order`, `turn_lanes`, `offset`, `width`, `fork_type`, geometry

Mid results (data/output/mid/):

File	Contents
`s3_fork_aligned.geojson`	Fork lane centerlines after start-node alignment
`s3_fork_buffered.geojson`	Buffered fork polygons
`s3_cycle_lines.geojson`	Cycleway centerlines after endpoint alignment
`s3_cycle_polys.geojson`	Cycleway polygons after alignment
`s3_pre_bifork_lines.geojson`	Lane centerlines after fork+cycle+accessor fixes, pre-bi-fork pass
`s3_pre_bifork_polys.geojson`	Same state, polygons
`s3_final_lane_lines.geojson`	Final lane centerlines — all passes complete
`s3_final_lane_polys.geojson`	Final lane polygons
`s3_final_boundaries.geojson`	Final road boundary lines

Stage 4 — Post-Process Road Polygons

Inputs: data/input/delft_osm_ways.geojson + s1_road_attributes.geojson + s3_lane_segments.geojson

degree_true (corrected intersection degree for interior nodes) is computed directly from the road graph geometry — no additional external file required.

Stage results → stage 5:

File	Rows	Role
`data/output/stage/s4_network_nodes.geojson`	408	Network nodes with `degree` (momepy count) and `degree_true` (corrected for interior nodes). Stage 5 overrides `degree` with `degree_true` where set.
`data/output/stage/s4_road_polygons.geojson`	395	Per-road polygons after all trim passes. Stage 5 clips these by intersection polygons.

Mid results (data/output/mid/):

File	After which trim pass
`s4_trim_degree2.geojson`	Degree-2 pseudo-intersection node trim
`s4_trim_bikelanes.geojson`	Mixed vehicle+cycleway boundary trim
`s4_fixed_polys.geojson`	Merged state after pseudo + mixed trims
`s4_endpoint_trim.geojson`	Degree-3 dead-end stub trim
`s4_fixed_polys2.geojson`	State after endpoint trim
`s4_multitype_trim.geojson`	Final multi-type trim (cycle + vehicle at same node)

Stage 5 — Generate Intersection Polygons

Inputs: data/input/delft_osm_ways.geojson + s4_road_polygons.geojson + s4_network_nodes.geojson + s3_lane_segments.geojson

Final outputs:

File	Rows	Contents
`data/output/road_intersections.geojson`	108	Intersection polygons (vehicle and cycleway, including roundabouts), each buffered 0.005 m
`data/output/road_segments.geojson`	329	Road polygons with intersections removed; `type`, `width`, `lanes`, `fork_type`, `bridge`, `tunnel`, `junction`, `name`, `oneway`

Mid results: none.

Output types

Type	Location	Description	Can delete?
Stage result	`data/output/stage/`	Required input for the next stage	No, until downstream stage completes
Mid result	`data/output/mid/`	Inspection snapshots for debugging	Yes, after verification
Final output	`data/output/`	End product of the pipeline	Keep

Project structure

ConRE-OSM/
├── run.py                  ← CLI entry point; all paths defined here
├── README.md
├── LICENSE                 ← GPL-3.0-or-later (code)
├── LICENSE-DATA            ← CC BY-NC 4.0 (output data)
├── requirements.txt
├── pipeline/
│   ├── stage1_centerlines.py
│   ├── stage2_lane_offset.py
│   ├── stage3_lane_geometry.py
│   ├── stage4_road_polygon.py
│   └── stage5_intersections.py
└── data/
    ├── input/
    │   └── delft_osm_ways.geojson   ← place your OSM export here
    └── output/
        ├── stage/                   ← stage results (auto-created)
        ├── mid/                     ← mid results (auto-created)
        ├── road_intersections.geojson
        └── road_segments.geojson

Licenses

Code — GNU General Public License v3.0 or later (GPL-3.0-or-later) See LICENSE · Full text

Output data — Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) See LICENSE-DATA · Full text

Input data — © OpenStreetMap contributors, Open Database License (ODbL 1.0) Any use of the output data must retain OSM attribution.

Navigation Menu

Repository files navigation

ConRE-OSM

OSM Road Surface Reconstruction — Case Study: Delft

Citation

Thesis

Tool

References

Methods and tools used

Related projects

Final outputs

Setup

Run

Pipeline overview

Data flow — detailed

Stage 1 — Extract, Classify and Label Road Centerlines

Stage 2 — Offset Lane Centerlines

Stage 3 — Modify Lane Geometry for Topology

Stage 4 — Post-Process Road Polygons

Stage 5 — Generate Intersection Polygons

Output types

Project structure

Licenses

About

Releases

Packages

Contributors

Languages

Name		Name	Last commit message	Last commit date
pipeline		pipeline
.gitignore		.gitignore
LICENSE		LICENSE
LICENSE-DATA		LICENSE-DATA
README.md		README.md
requirements.txt		requirements.txt
run.py		run.py

Latest commit

History

Folders and files

Repository files navigation

ConRE-OSM

OSM Road Surface Reconstruction — Case Study: Delft

Citation

Thesis

Tool

References

Methods and tools used

Related projects

Final outputs

Setup

Run

Pipeline overview

Data flow — detailed

Stage 1 — Extract, Classify and Label Road Centerlines

Stage 2 — Offset Lane Centerlines

Stage 3 — Modify Lane Geometry for Topology

Stage 4 — Post-Process Road Polygons

Stage 5 — Generate Intersection Polygons

Output types

Project structure

Licenses

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages