Skip to main content

Processing Level

This guide summarizes common hyperspectral processing levels and how they relate to datasets in Clarity.

Overview​

  • Raw: Sensor digital numbers (DN). Uncorrected, includes sensor/system effects. Not supported for dataset creation.
  • Dark: Dark-frame corrected DN to remove sensor offset. Still not radiometrically calibrated. Not supported for dataset creation.
  • Radiance: Physically meaningful radiance values (e.g., W·sr⁻¹·m⁻²·nm⁻¹) after calibration. Supported for dataset creation. When training models on Radiance datasets, normalization should be applied to account for illumination differences.
  • Reflectance: Unitless surface reflectance after atmospheric/illumination correction (ratio of reflected to incoming energy). Supported for dataset creation.

Choosing a level​

  • Prefer Reflectance when illumination/atmospheric variation is significant or you need physically comparable spectra across scenes.
  • Use Radiance when reflectance conversion is unavailable or you want full control of normalization during model training.

See also: Creating datasets and Center wavelengths.