Data Source Enrichment

Enrich LiDAR point clouds with authoritative geographic data from IGN and government databases.

---

🎯 Overview

The IGN LiDAR HD package can enrich point clouds with ground truth data from multiple authoritative sources:

• BD TOPO® - IGN topographic database (buildings, roads, water, etc.)
• BD Forêt® - IGN forest database (tree species, forest types)
• RPG - Agricultural parcel register (crop types, land use)
• Cadastre - Land parcel boundaries and ownership

This enrichment significantly improves classification accuracy, especially for ASPRS classification.

---

🗺️ BD TOPO® (IGN Topographic Database)

France's reference topographic database, maintained by IGN (Institut National de l'Information Géographique et Forestière).

Available Features

Feature	ASPRS Code	Description
Buildings	6	Building footprints and heights
Roads	11	Road networks with classifications
Railways	10	Railway tracks and stations
Water	9	Rivers, lakes, water bodies
Vegetation	5	Parks, forests, green spaces
Bridges	17	Bridge structures
Parking	40	Parking lots and areas
Sports	41	Sports facilities
Cemeteries	42	Cemetery boundaries
Power Lines	43	Electrical transmission lines

Configuration

data_sources:
  # Enable BD TOPO
  bd_topo_enabled: true

  # Select specific features
  bd_topo_buildings: true # ASPRS code 6
  bd_topo_roads: true # ASPRS code 11
  bd_topo_water: true # ASPRS code 9
  bd_topo_vegetation: true # ASPRS code 5

  # Cache settings
  bd_topo_cache_dir: "cache/bd_topo"

Usage Example

ign-lidar-hd process \
    --config-file configs/default_v3.yaml \
    data_sources.bd_topo_enabled=true \
    data_sources.bd_topo_buildings=true \
    data_sources.bd_topo_roads=true

Data Quality

• Accuracy: ± 1-5 meters planimetric, ± 1-3 meters altimetric
• Update frequency: Annual to quarterly (varies by region)
• Coverage: Complete coverage of metropolitan France
• Source: Aerial photography, satellite imagery, field surveys
• Version: BD TOPO® V3 (latest as of October 2025)

What BD TOPO Adds

For Buildings:

• Precise footprint geometry
• Building height information
• Building type classification
• Construction date (when available)

For Roads:

• Road centerlines and widths
• Road classification (highway, street, path)
• Surface type information
• Traffic direction

For Water:

• Water body boundaries
• Flow direction for rivers
• Water body type (lake, river, canal)
• Seasonal vs permanent

---

🌲 BD Forêt® (IGN Forest Database)

Detailed forest inventory database with tree species and forest characteristics.

Available Information

• Tree species: Oak, beech, pine, spruce, etc.
• Forest type: Deciduous, coniferous, mixed
• Density: Open, medium, dense
• Age class: Young, mature, old-growth
• Management: Production, conservation, recreation

Configuration

data_sources:
  bd_foret_enabled: true
  bd_foret_cache_dir: "cache/bd_foret"

Usage Example

ign-lidar-hd process \
    --config-file configs/default_v3.yaml \
    data_sources.bd_foret_enabled=true

What BD Forêt Adds

• Vegetation classification: Distinguishes tree species
• Forest structure: Height classes, density
• Ecological zones: Protected areas, biodiversity hotspots
• Forest management: Production vs conservation areas

Data Quality

• Accuracy: Forest boundary ± 10 meters, species identification ~90%
• Update frequency: Every 5-10 years
• Coverage: Forest areas in metropolitan France
• Source: Aerial photography, field surveys, satellite imagery
• Version: BD Forêt® V2 (latest as of October 2025)

---

🌾 RPG (Registre Parcellaire Graphique)

Agricultural parcel register maintained by the French Ministry of Agriculture.

Available Information

• Crop types: 28+ crop categories
• Land use: Agricultural, grassland, fallow
• Parcel boundaries: Field-level precision
• Cultivation practices: Organic, conventional
• CAP subsidies: EU Common Agricultural Policy data

Configuration

data_sources:
  rpg_enabled: true
  rpg_year: 2024 # Use 2024 data
  rpg_cache_dir: "cache/rpg"

Usage Example

ign-lidar-hd process \
    --config-file configs/default_v3.yaml \
    data_sources.rpg_enabled=true \
    data_sources.rpg_year=2024

Crop Categories (Examples)

• Cereals (wheat, barley, corn)
• Oilseeds (rapeseed, sunflower)
• Protein crops (soybeans, peas)
• Root crops (potatoes, beets)
• Forage crops (alfalfa, grass)
• Permanent crops (vineyards, orchards)

What RPG Adds

• Agricultural land classification: Precise crop type identification
• Seasonal context: Current year's cultivation
• Field boundaries: Parcel-level geometry
• Land use patterns: Crop rotation analysis

Data Quality

• Accuracy: Parcel boundary ± 5 meters, crop type ~95%
• Update frequency: Annual
• Coverage: Agricultural areas in metropolitan France
• Source: Farmer declarations, satellite verification
• Available years: 2020-2024

---

🏘️ Cadastre (Land Parcel Registry)

Official land parcel boundaries and ownership information.

Available Information

• Parcel boundaries: Legal property limits
• Parcel identifiers: Unique cadastral codes
• Land use type: Urban, agricultural, forest
• Building footprints: Building positions within parcels
• Section codes: Cadastral section identifiers

Configuration

data_sources:
  cadastre_enabled: true
  cadastre_cache_dir: "cache/cadastre"

Usage Example

ign-lidar-hd process \
    --config-file configs/default_v3.yaml \
    data_sources.cadastre_enabled=true

What Cadastre Adds

• Property boundaries: Precise parcel limits
• Building context: Buildings within land parcels
• Urban/rural distinction: Land use classification
• Parcel grouping: Group points by property

Data Quality

• Accuracy: Parcel boundary ± 0.5-2 meters (urban areas)
• Update frequency: Continuous updates
• Coverage: Complete coverage of France
• Source: Official land registry, surveyor measurements
• Maintained by: Direction Générale des Finances Publiques (DGFiP)

---

🎯 Combining Data Sources

Recommended Combinations

For ASPRS Classification (Best Accuracy)

data_sources:
  bd_topo_enabled: true
  bd_topo_buildings: true
  bd_topo_roads: true
  bd_topo_water: true
  bd_topo_vegetation: true
  bd_foret_enabled: true
  rpg_enabled: true
  rpg_year: 2024
  cadastre_enabled: true

Benefits:

• Maximum classification accuracy
• Complete ground truth coverage
• Cross-validation between sources

For Urban Analysis

data_sources:
  bd_topo_enabled: true
  bd_topo_buildings: true
  bd_topo_roads: true
  cadastre_enabled: true

Benefits:

• Building and infrastructure focus
• Property boundary context
• Road network integration

For Forest/Rural Analysis

data_sources:
  bd_foret_enabled: true
  rpg_enabled: true
  rpg_year: 2024

Benefits:

• Vegetation and agriculture focus
• Tree species identification
• Crop type mapping

---

📊 Performance Impact

Processing Time

Data Sources	Relative Time	Cache Benefit
None	1.0x	-
BD TOPO only	1.2x	Yes (3x faster after cache)
BD TOPO + BD Forêt	1.4x	Yes (4x faster after cache)
All sources	1.8x	Yes (5x faster after cache)

File Size Impact

Data source enrichment does not significantly increase file size:

• Point cloud size: No change (same points)
• Classification field: Already present in LAZ
• Metadata: Minimal increase (<1%)

Caching System

All data sources use intelligent caching:

# Cache configuration
data_sources:
  bd_topo_cache_dir: "cache/bd_topo"
  bd_foret_cache_dir: "cache/bd_foret"
  rpg_cache_dir: "cache/rpg"
  cadastre_cache_dir: "cache/cadastre"

Cache benefits:

• First run: Downloads and caches data
• Subsequent runs: Uses cached data (5x faster)
• Automatic updates: Checks for new data versions
• Disk space: ~100-500 MB per region

---

🔍 How It Works

Data Fetching Process

1. Query: System identifies tile bounding box
2. WFS Request: Queries IGN/government Web Feature Service
3. Filter: Retrieves only relevant features
4. Cache: Stores data locally for reuse
5. Intersect: Identifies points within feature geometries
6. Classify: Assigns ASPRS codes based on feature type

Spatial Operations

# Pseudo-code for enrichment
for tile in tiles:
    # Fetch relevant features
    buildings = bd_topo.get_buildings(tile.bbox)
    roads = bd_topo.get_roads(tile.bbox)

    # Classify points
    for point in tile.points:
        if point.intersects(buildings):
            point.classification = 6  # Building
        elif point.intersects(roads):
            point.classification = 11  # Road

Priority Rules

When a point intersects multiple features:

1. Buildings (code 6) - Highest priority
2. Water (code 9) - High priority
3. Roads (code 11) - Medium priority
4. Vegetation (code 5) - Lower priority
5. Ground (code 2) - Default

---

🚀 Quick Start Examples

Example 1: Basic Enrichment

ign-lidar-hd process \
    input_dir=data/raw \
    output_dir=data/enriched \
    data_sources.bd_topo_enabled=true

Example 2: Full ASPRS Classification

ign-lidar-hd process \
    --config-file configs/presets/asprs_classification.yaml \
    input_dir=data/raw \
    output_dir=data/asprs

Example 3: Custom Feature Selection

ign-lidar-hd process \
    input_dir=data/raw \
    output_dir=data/custom \
    data_sources.bd_topo_enabled=true \
    data_sources.bd_topo_buildings=true \
    data_sources.bd_topo_roads=true \
    data_sources.bd_topo_water=false \
    data_sources.bd_foret_enabled=false

---

📚 See Also

• Configuration System v3.0
• Feature Modes Guide
• ASPRS Classification Reference
• Processing Modes

---

❓ FAQ

Do I need all data sources?

No. Enable only what you need:

• ASPRS classification: BD TOPO + BD Forêt recommended
• Building analysis: BD TOPO buildings + Cadastre
• Forest analysis: BD Forêt only
• Agriculture: RPG only

How much does caching help?

First run: Downloads data (slower)
Subsequent runs: Uses cache (5x faster)

Can I use my own data sources?

Yes! See the Custom Data Sources guide.

What if WFS is unavailable?

The system falls back to geometry-based classification using LiDAR features only.

How accurate is the classification?

With data sources: 85-95% accuracy
Without data sources: 70-80% accuracy

The improvement is significant, especially for infrastructure classes.