Multispectral imaging captures a scene in several discrete wavelength bands — visible colours plus infrared ones — so each pixel becomes a spectral signature. Vegetation health, water, minerals and burn scars all reveal themselves in band combinations invisible to the eye.
Understanding Multispectral Imaging
From photons to products
Raw multispectral data crosses a processing ladder: radiometric calibration to physical radiance, atmospheric correction stripping haze and scattering to surface reflectance, orthorectification onto map geometry — then analysis: band indices, classification into land-cover maps, and change detection between revisits. Modern pipelines run the ladder automatically at continental scale, turning weekly imagery into crop forecasts, deforestation alerts and flood extents within hours of downlink.
Choosing bands is choosing questions
Every band placement encodes intent: red-edge channels (~700–750 nm) sharpen crop-stress detection; shortwave-infrared separates snow from cloud and reads soil moisture and burn severity; thermal bands measure surface temperature for irrigation and fire mapping; coastal-blue penetrates shallow water. Sensor designers arrange a spectral toolkit per mission, and analysts' band-combination lore — which triplet shows floods, which shows smoke through haze — is the field's working craft.