Land Cover Classification

Land cover classification uses remote sensing to map what occupies the Earth’s surface, supporting environmental monitoring, agriculture, conservation, and urban planning. Common approaches include random forests and other classical classifiers, CNN and transformer segmentation models, and time-series methods that use seasonal changes; newer systems also combine optical, radar, hyperspectral, and LiDAR data or use pretrained geospatial embeddings. A central challenge is transferring maps across regions, years, sensors, and climate zones, where imagery and class definitions can differ. Researchers also weigh accuracy against annotation needs and compute, while addressing rare or spectrally similar classes and validating maps with spatially independent data.