PROJECTS_HUB
// SYSTEM_ID: RESEARCH_05 // STATUS: ACTIVE_RESEARCH // SENTINEL-2

GeoAI
Research Lab

Applied artificial intelligence for territorial analysis — Sentinel-2 satellite change detection, machine learning land cover classification, and predictive models for cadastral quality scoring.

94%
Detection Accuracy
2.3km²
Urban Change Detected
847
Polygons Vectorized
10m
Spatial Resolution
01
Urban Change Detection
Multitemporal Sentinel-2 imagery analysis comparing 2020 vs 2025. Random Forest classifier trained on labeled urban expansion samples from IGAC datasets.
Sentinel-2RasterioScikit-learnNumPy
02
Cadastral Quality ML Predictor
Gradient Boosting model predicting topological error probability in predios based on neighborhood, area, shape index, and historical QA results. Achieves 89% precision.
XGBoostPostGISGeoPandasSHAP
03
Land Cover Classification
4-class supervised classification (urban, vegetation, water, bare soil) over Medellín metropolitan area using spectral indices NDVI, NDWI, MNDWI from Sentinel-2 bands.
NDVIRandom ForestShapelyGDAL
04
Drone + Satellite Fusion
Integration of centimeter-resolution drone orthophotos with 10m Sentinel data for hybrid cadastral update workflows. Reduces field verification trips by ~60%.
AgisoftOpenDroneMapRasterioQGIS
LIVE GEOAI DEMO BACK TO MAIN