ArcheoVLM
Phase 5: Georeferencing & Multi-Modal Verification
Convert pixel-based findings into geographic data and assemble comprehensive verification packages
Convert pixel-based findings into real-world geographic data and automatically assemble a rich, multi-modal verification package for each potential archaeological site.
Coordinate Transformation
Convert YOLO bounding box coordinates to WGS84 latitude/longitude using GeoTIFF metadata
Spatial Accuracy
Maintain high spatial accuracy through proper coordinate system transformations
Optical Satellite Imagery
Recent, low-cloud-cover imagery with True-Color, NDVI, and NDMI products
SAR Imagery
L-band SAR data to detect subsurface soil moisture anomalies indicating buried earthworks
Historical Text Analysis
Two-stage toponym resolution pipeline to geoparse historical Amazonian texts and extract relevant archaeological mentions
Historical Aerial Photography
Historical aerial photos from Brazilian and international archives
LiDAR Products
Satellite Data
Historical Context
AI Analysis
Processing Steps:
- • Georeferencing: Convert all pixel coordinates to WGS84 coordinates
- • Data Acquisition: Query external APIs and archives for each site location
- • Image Processing: Georeference and crop relevant imagery
- • Text Mining: Extract and correlate historical text references
- • Package Assembly: Compile all data layers into structured verification packages
Master Site Database
potential_sites.geojson with all georeferenced discoveries
Verification Packages
Complete multi-modal data packages for each high-confidence site