geofeatures 0.4.0

Zonal statistics and NDVI feature extraction for remote sensing ML pipelines

GeoFeatures

Python library for spectral index computation, zonal statistics, and vector/raster operations for remote sensing ML pipelines.

Install

pip install geofeatures

Spectral indices

Function	Detects	Inputs
compute_ndvi()	Vegetation health	NIR, Red
compute_evi()	Dense vegetation	NIR, Red, Blue
compute_savi()	Sparse vegetation	NIR, Red
compute_gndvi()	Chlorophyll	NIR, Green
compute_ndwi()	Surface water	Green, NIR
compute_mndwi()	Urban water	Green, SWIR
compute_ndmi()	Vegetation moisture	NIR, SWIR
compute_ndbi()	Built-up areas	SWIR, NIR
compute_bsi()	Bare soil	SWIR, Red, NIR, Blue
compute_nbr()	Burn severity	NIR, SWIR2

Zonal statistics

• extract_zonal_features(raster_path, vector_gdf, stats)
• ndvi_zonal_stats(...) — convenience wrapper (NDVI + zonal stats in one call)

Vector operations

• merge_shapefiles(gdf_list, target_crs=None) — merge GeoDataFrames, handling CRS mismatches
• dissolve_by_attribute(vector_gdf, attribute, agg_func="first") — dissolve polygons by shared attribute
• clip_vector(vector_gdf, clip_boundary_gdf) — clip vector to a boundary

Raster-vector operations

• clip_raster_by_vector(raster_path, vector_gdf, output_path) — clip a raster to a vector boundary

Validated

Tested end-to-end on real Sentinel-2 L2A imagery over Ogun State, Nigeria. 18 unit tests with hand-verified formula outputs and edge-case coverage (CRS mismatches, missing columns, spatial overlap logic).

License

MIT

Visualization

• plot_raster(raster_path, title, cmap) — quick raster preview
• plot_vector(vector_gdf, column, title) — quick vector/choropleth preview
• plot_zonal_result(vector_gdf, column, title) — convenience wrapper for zonal stats results

Format conversion

• convert_vector_format(input_path, output_path) — convert between Shapefile, GeoJSON, GPKG, KML
• load_kmz(path) — load a KMZ (zipped KML) file
• load_vector(path, target_crs, fix_invalid) — load with automatic geometry repair
• load_raster_as_array(path) — load raster as numpy array + metadata