azavea / nasa-hyperspectral

An event-driven image processing pipeline for developing our foundational capability to work with HSI data sources.
15 stars 3 forks source link

Chemical Cloud Detection #107

Closed pomadchin closed 2 years ago

pomadchin commented 3 years ago

Given that we now have much better context regarding hyperspectral analysis, there appears to be three main analytical techniques that, if we master them, will allow us to tackle a wide array of different problems:

  1. Unmixing
  2. Target detection
  3. Plume detection

This epic deals with work related to plume detection. This is applicable for many topics including methane detection, shipping lane identification (by identifying sulfur compounds produced in the combustion of diesel fuel), and identification of plastics incineration sites, the latter of which is the motivating application for pursuing this area.

The underlying model for plume detection shares features with atmospheric correction, as they are both radiative transfer problems, though it's not clear if we need to solve an atmospheric correction-style problem to do plume detection. Further investigations will be required.

jpolchlo commented 3 years ago

Incidentally, we may want to consider restructuring the NASA status reports to reflect the understanding that techniques are king, and perhaps should take precedence over specific applications. We have both target detection and plume detection examples; we ought to look for a compelling application of unmixing as well. Thoughts, @echeipesh @jcahail ?