Green Area Inventory / Grünflächeninventar

Caution

This is an unofficial mockup for demonstration purposes only. All data is fictional. Not all features are fully functional. This project serves as a visual and conceptual prototype — it is not intended for production use.

Simple Viewer: https://bbl-dres.github.io/green-inventory

Advanced Prototype: https://bbl-dres.github.io/green-inventory/prototype1

Interactive GIS web application mockup for urban green space inventory, maintenance planning, and field survey — built around interactive maps, care profiles, and task management.

Project structure

green-inventory/
├── index.html              # Map viewer (MapLibre GL JS + CARTO basemap)
├── extract_features.py     # PDF → GeoJSON extraction script
├── data/
│   ├── [838147959] 1602.GR_Mühlestrasse 2+4+6+8Grünflächenpflege.pdf
│   └── [838147959] 1602.GR_Mühlestrasse 2+4+6+8Grünflächenpflege.geojson
└── prototype1/             # Earlier prototype (separate app)

extract_features.py

Extracts vector features from a Swiss landscape plan PDF (Grünflächenpflege, scale 1:650) and outputs a WGS 84 GeoJSON file.

Dependencies

pip install pymupdf shapely numpy pyproj

Usage

python extract_features.py

Input and output paths are configured at the top of the script (PDF_PATH, OUTPUT_PATH).

How it works

The script reads all vector drawing paths from the PDF page using PyMuPDF and processes them in two passes:

Pass 1 — Large filled areas (`>= 20 pt` in both dimensions)

These are solid-colour polygons and hatching stripes that define feature boundaries directly. Paths are grouped by their classified feature type, then merged using a buffer → union → erode operation (MERGE_BUFFER_M = 1.8 m) to collapse individual hatch stripes into unified area polygons. The result is exploded back into individual polygons per spatially disconnected region.

Pass 2 — Pattern tiles (`< 20 pt`)

These are the small repeated elements (dots, dashes, crosshatch marks) that fill patterned areas. Each tile's bounding rect is converted to a small Shapely polygon, then the same buffer → union → erode approach merges adjacent tiles into area polygons. This replaces the earlier convex-hull-of-centroids approach, which over-estimated boundaries.

Color classification

Each path's fill RGB is matched against FEATURE_COLORS (a lookup table of ~30 entries) using Euclidean distance in RGB space with a tolerance of 0.04. Unclassified fills (white building footprints, etc.) are skipped — those will be sourced from official survey data.

Georeferencing

A single ground control point (GCP) is used:

GCP	WGS 84	Assumed local position
1	46.97510 N, 7.47417 E	Map centre

The scale is calibrated from the plan's scale bar (174.49 PDF points = 40 m → 1 pt ≈ 0.2292 m). Local metre coordinates are converted to WGS 84 via Swiss LV95 (EPSG:2056) as an intermediate CRS for accurate metric offsets.

Note: North is assumed to be up. The plan may have a slight rotation. Use the Position bearbeiten tool in the map viewer to drag and correct the position, then save the updated GeoJSON.

Legend layers

Building footprints and the site boundary (red outline) are intentionally excluded — these will be sourced from official Swiss survey data (amtliche Vermessung).

How to read the matching column:

Solid fill — one or more large closed polygons drawn with this fill colour. Directly converted to geometry.
Hatch stripes — many overlapping diagonal parallelograms in this colour tiling the area. Merged via buffer→union→erode.
Pattern tiles — thousands of tiny (< 5 mm²) repeated dot/dash/crosshatch elements. Merged the same way.
⚠️ Cannot distinguish — two or more legend subtypes share an identical fill colour; they are merged into a single subtype in the output.
➖ Not present — this legend type does not appear in this particular plan file.

Rasen

Subtype	Category	PDF colour	Match method
Geb.Rasen kf.	`lawn`	`#97e600` rgb(0.596, 0.902, 0)	Solid fill
Blumenrasen	`lawn`	`#97e600` on white	⚠️ Same base colour as Geb.Rasen kf. — dot tiles not yet distinguished; merged into `lawn`

Wiesen

Subtype	Category	PDF colour	Match method
Feuchtwiese	`meadow`	`#a8e3d9` rgb(0.659, 0.890, 0.851)	➖ Not present in this plan
Blumenwiese gf.	`meadow`	`#e9ffbd` rgb(0.914, 1.0, 0.745)	⚠️ Same colour as Blumenwiese kf. — stripe overlay tiles not yet distinguished; both merged into `Blumenwiese`
Blumenwiese kf.	`meadow`	`#e9ffbd` rgb(0.914, 1.0, 0.745)	Solid fill
Magerrasen	`meadow`	`#f5f579` rgb(0.961, 0.961, 0.478)	⚠️ Same base colour as Saumvegetation — dot tile colours differ but not yet classified separately; merged into `Saumvegetation`
Saumvegetation	`meadow`	`#f5f579` rgb(0.961, 0.961, 0.478)	Solid fill

Rabatten

Subtype	Category	PDF colour	Match method
Wechselflor	`planting_bed`	`#ff73df` rgb(1.0, 0.451, 0.875)	➖ Not present in this plan
Moorbeet	`planting_bed`	`#5c45a8` rgb(0.361, 0.271, 0.659)	➖ Not present in this plan
Stauden. int.	`planting_bed`	`#df73ff` rgb(0.875, 0.451, 1.0)	⚠️ Same base colour as Stauden ext. — stripe tiles not yet distinguished; both merged into `Stauden`
Stauden. ext.	`planting_bed`	`#df73ff` rgb(0.875, 0.451, 1.0)	Solid fill
Beetrosen	`planting_bed`	`#ff5500` rgb(1.0, 0.333, 0)	➖ Not present in this plan
Ruderalfläche	`planting_bed`	`#ffaa00` rgb(1.0, 0.667, 0)	➖ Not present in this plan

Hecken

Subtype	Category	PDF colour	Match method
Wildhecke	`hedge`	`#d9d89e` rgb(0.843, 0.843, 0.620)	➖ Not present in this plan
Formhecke + 1.5 m	`hedge`	`#a86f00` rgb(0.659, 0.439, 0)	⚠️ Same tile colour as Formhecke - 1.5 m — height cannot be read from fill alone; merged into `Formhecke`
Formhecke - 1.5 m	`hedge`	`#a86f00` rgb(0.659, 0.439, 0)	Pattern tiles (small hatch marks)

Gehölzflächen

Subtype	Category	PDF colour	Match method
Gehölz & Bodend.	`woody_area`	`#896e44` rgb(0.537, 0.439, 0.267)	➖ Not present in this plan
Bodendecker	`woody_area`	`#896e44` rgb(0.537, 0.439, 0.267)	➖ Not present in this plan
Gehölzrabatte	`woody_area`	`#718844` rgb(0.447, 0.537, 0.267)	Pattern tiles (small dots)
Wald	`woody_area`	`#267300` rgb(0.150, 0.450, 0)	➖ Not present in this plan

Spezielle Bepflanzungsformen

Subtype	Category	PDF colour	Match method
Dach: ext. Stauden	`special_planting`	`#ffbdbd` rgb(1.0, 0.745, 0.745)	Solid fill

Beläge

Subtype	Category	PDF colour	Match method
Asphaltbelag	`surface`	`#686868` rgb(0.408, 0.408, 0.408)	➖ Not present in this plan — the dark-grey tiles (`#4e4e4e`) are 12-sided circle polygons (Bollensteine), not asphalt texture
Rasengittersteine	`surface`	`#6fa800` rgb(0.439, 0.659, 0)	Pattern tiles (green diamond crosshatch, ~1 760 tiles)
Holzhäckselbelag	`surface`	`#a83800` rgb(0.659, 0.220, 0)	Pattern tiles (dark-brown diamond crosshatch, ~7 500 tiles)
Chaussierung	`surface`	`#cd8866` rgb(0.804, 0.537, 0.4)	Fill+stroke boundaries (17 direct polygon outlines)
Betonpl./Verbund/Naturstein	`surface`	`#9c9c9c` rgb(0.612, 0.612, 0.612)	➖ Not present in this plan
Geröllstreifen/Bollensteine	`surface`	`#4e4e4e` rgb(0.306, 0.306, 0.306)	Pattern tiles (dark-grey 12-gon circle polygons, ~2 280 tiles) — previously misidentified as Asphaltbelag

Wasserflächen

Subtype	Category	PDF colour	Match method
Brunnen	`water`	`#00a9e6` rgb(0, 0.663, 0.902)	➖ Not present in this plan
Gewässer, ruhend	`water`	`#006fff` rgb(0, 0.439, 1.0)	➖ Not present in this plan

Anderes

Subtype	Category	PDF colour	Match method
Anderes	`other`	`#ffff00` rgb(1.0, 1.0, 0)	➖ Not present in this plan

Known limitations

Issue	Affected subtypes	Workaround
Identical base fill colour	Blumenwiese gf. vs kf. / Stauden int. vs ext. / Formhecke +1.5m vs -1.5m	The distinguishing feature is the overlay pattern colour, not the base fill. Future improvement: cross-reference tile colour with base polygon area to resolve.
Magerrasen vs Saumvegetation	Both use `#f5f579` base	Tile dot colour differs (yellow vs orange-brown) — could be classified separately with a lower `COLOR_TOL`.
Gehölz & Bodend. vs Bodendecker	Both use `#896e44`	These differ only by the presence/absence of dot overlay tiles.
Formhecke only partially in map	8 tiles at the east edge (x ≈ MAP_X_MAX)	The hedge runs mostly outside the plan extent; the 8 edge tiles are too small to produce a meaningful polygon after merging.
Plan-specific presence	Many types absent	Only the feature types drawn in a given PDF will appear. Run the script on a different plan to get different results.

GeoJSON output format

{
  "type": "FeatureCollection",
  "crs": { "type": "name", "properties": { "name": "urn:ogc:def:crs:OGC:1.3:CRS84" } },
  "metadata": {
    "source_pdf": "...",
    "site": "Muehlestrasse 2-6, 3063 Ittigen",
    "scale": "1:650",
    "gcp1_wgs84": [46.97510, 7.47417],
    "map_extent_m": [245.7, 251.8],
    "approx_bbox_wgs84": [7.4726, 46.9740, 7.4758, 46.9762],
    "offset_m": [0.0, 0.0]
  },
  "features": [
    {
      "type": "Feature",
      "geometry": { "type": "Polygon", "coordinates": [...] },
      "properties": {
        "feature_type": "Belag",
        "subtype": "Chaussierung",
        "category": "surface",
        "fill_rgb": [0.804, 0.537, 0.4],
        "source": "area_merged",
        "area_m2": 283.0
      }
    }
  ]
}

source field values:

area_merged — reconstructed from large filled paths (solid areas or merged hatch stripes)
pattern_merged — reconstructed from small pattern tiles (dots, dashes, crosshatch)

Map viewer

Open index.html via a local HTTP server (required for the fetch call):

python -m http.server 8000
# then open http://localhost:8000

Features:

CARTO Positron basemap via MapLibre GL JS (no API key required)
Full legend panel matching the plan legend, grouped identically
Eye icon toggles per legend group
Hover popup with feature type, subtype, area m²
Position bearbeiten — drag the entire layer to correct georeferencing offset, then download the updated GeoJSON

Name		Name	Last commit message	Last commit date
Latest commit History 68 Commits
.claude		.claude
assets		assets
css		css
data		data
js		js
prototype1		prototype1
README.md		README.md
extract_features.py		extract_features.py
index.html		index.html

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Green Area Inventory / Grünflächeninventar

Project structure

extract_features.py

Dependencies

Usage

How it works

Pass 1 — Large filled areas (`>= 20 pt` in both dimensions)

Pass 2 — Pattern tiles (`< 20 pt`)

Color classification

Georeferencing

Legend layers

Rasen

Wiesen

Rabatten

Hecken

Gehölzflächen

Spezielle Bepflanzungsformen

Beläge

Wasserflächen

Anderes

Known limitations

GeoJSON output format

Map viewer

About

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

Green Area Inventory / Grünflächeninventar

Project structure

extract_features.py

Dependencies

Usage

How it works

Pass 1 — Large filled areas (>= 20 pt in both dimensions)

Pass 2 — Pattern tiles (< 20 pt)

Color classification

Georeferencing

Legend layers

Rasen

Wiesen

Rabatten

Hecken

Gehölzflächen

Spezielle Bepflanzungsformen

Beläge

Wasserflächen

Anderes

Known limitations

GeoJSON output format

Map viewer

About

Topics

Resources

Uh oh!

Stars

Watchers

Forks

Contributors

Uh oh!

Languages

Pass 1 — Large filled areas (`>= 20 pt` in both dimensions)

Pass 2 — Pattern tiles (`< 20 pt`)