An overview of applying AI to reduce noise in planetary hyperspectral observations and improve mineral interpretation.
Hyperspectral instruments capture detailed information about the composition of planetary surfaces, but sensor degradation, noise and limited ground truth can restrict interpretation. Machine learning provides a way to recover structure from those observations without treating every artefact as a meaningful signal.
Working with academic researchers, GeonatIQ has explored denoising approaches for CRISM data from Mars. The aim is to improve the fidelity of mineralogical signatures and make large orbital datasets more useful for scientific investigation.
The same principles can support classification, anomaly detection and automated mineral mapping across lunar, asteroid and other planetary missions.