Logo Lanfrica

ibrahimadiatta/Sentinel2-LandCover-Classification

Domaine:

geospatial

Type de record:

project
Créateur:
ibr
Hôte:
Land cover classification of Thiès Nord (Senegal) using Sentinel-2 imagery and Google Earth Engine. # Land Cover Classification using Sentinel-2 and Google Earth Engine ## Description This project performs supervised land cover classification over Thiès Nord (Senegal) using Sentinel-2 imagery and the CART algorithm available in Google Earth Engine. ## Study Area Thiès Nord, Senegal. ## Dataset - Sentinel-2 Surface Reflectance - Period: - May 2023 - September 2023 ## Spectral Bands - B2 - B3 - B4 - B8 - B11 - B12 ## Classes | Code | Class | |------|--------------------| | 1 | Vegetation | | 2 | Bare Soil | | 3 | Buildings | | 4 | Agricultural Soil | | 5 | Concrete | | 6 | Roads | ## Methodology 1. Import Sentinel-2 imagery 2. Filter by date 3. Filter by study area 4. Compute median composite 5. Create training samples 6. Train CART classifier 7. Classify image 8. Assess classification accuracy 9. Export GeoTIFF ## Results ### Sentinel-2 Composite ### Training Samples ### CART Classification ## Accuracy Assessment The classification performance was evaluated using an independent validation dataset. | Metric | Value | |--------|------:| | Overall Accuracy | 0.78 | | Kappa Coefficient | 0.72 | ### Producer's Accuracy | Class | Accuracy | |--------|---------:| | Vegetation | 0.85 | | Bare Soil | 0.76 | | Buildings | 0.69 | | Agricultural Soil | 0.98 | | Concrete | 0.70 | | Roads | 0.80 | ### User's Accuracy | Class | Accuracy | |--------|---------:| | Vegetation | 0.82 | | Bare Soil | 0.81 | | Buildings | 0.73 | | Agricultural Soil | 0.98 | | Concrete | 0.57 | | Roads | 0.82 | ## Confusion Matrix | Reference \\ Predicted | Veg | Bare Soil | Buildings | Agr. Soil | Concrete | Roads | |-------------------------|----:|----------:|----------:|----------:|---------:|------:| | Vegetation | 22 | 1 | 1 | 1 | 0 | 1 | | Bare Soil | 0 | 22 | 5 | 0 | 2 | 0 | | Buildings | 0 | 3 | 49 | 0 | 13 | 6 …