Geospatial feature extraction is the derivation of descriptive variables or discrete objects from georeferenced Earth observation data so that they can be classified, mapped or analysed.12
A remote sensing review of wetland classification places feature extraction and selection as a workflow step covering spectral features such as vegetation and water indices, spatial relationship features such as texture, and temporal sequence features.1 It reports multi-feature fusion as a key strategy and notes that convolutional networks extract local spatial features such as texture and shape.1 NASA defines Level 2 products as derived geophysical variables at the same resolution and location as the Level 1 source data, which is one product-level outcome of derivation.2
ESA Φ-lab says machine learning is critical to accelerating insight extraction from data while domain experts remain essential to interpret it.3 The retained sources do not give a general standard definition, and the review is a single wetland-focused witness. The term also covers two senses, raster variables and vector objects, that the sources do not separate.
Footnotes
-
Frontiers in Remote Sensing, “Remote sensing-based detailed wetland classification”. https://www.frontiersin.org/journals/remote-sensing/articles/10.3389/frsen.2026.1852249/full ↩ ↩2 ↩3
-
NASA Earthdata, “Data Processing Levels”. https://www.earthdata.nasa.gov/learn/earth-observation-data-basics/data-processing-levels ↩ ↩2
-
ESA Φ-lab, “Focus on AI4EO”. https://philab.esa.int/flagship-programmes/focus-on-ai4eo/ ↩