Stage 05 MESMA
DO NOT EDIT OUTSIDE MARKERS
Using the spectral library
The spectral library from Stage 04 provides the
candidate endmember spectra. You load it as a NumPy array and pass it to the
MESMA routine alongside the target image. MESMA iterates through library
combinations to find the model with the lowest root mean square error (RMSE).
from unmixing.el_mesma import MesmaCore
mesma = MesmaCore()
fractions, residuals = mesma._mesma(image, library)
Endmember selection strategies
- Exhaustive search – evaluate all combinations up to a fixed complexity.
- Class-based – restrict models to endmembers drawn from predefined
classes such as vegetation or soil.
- Random sampling – sample combinations to reduce runtime for large
libraries.
Outputs
- Per-endmember fraction maps showing the proportional contribution of each
material and a shade fraction.
- A residual raster capturing the difference between observed and reconstructed
spectra.
Validation
- Verify that the fractions for each pixel sum to approximately
1.0.
- Discard models with RMSE above a user-defined threshold to ensure a reliable
fit.
Last updated: 2025-08-18