Empirical Reflectance Retrieval


This page will discuss the research required to demonstrate the feasibility of directly retrieving reflectance information from Full Spectral Imaging data. Empirical Reflectance Retrieval will eliminate the need for modeling. Empirical Reflectance Retrieval will provide measurements that give researchers spectral reflectance at the target, as opposed to the current system that only provides a measurement of spectral radiance at the top of the atmosphere.

The basic tenet of Empirical Reflectance Retrieval is; "The more you know, the easier it is to figure out what you do not know".  It really boils down to an exercise in pattern recognition.  The idea is that the more validated information you acquire, both spectrally and spatially, the better you will be able to match new information with the previously acquired information.  This can be a completely automatic process.  High spectral and spatial resolution hyperspectral systems that acquire huge amounts of information are ideal for this approach.  It would also include auxiliary sources of information, both spectral and spatial.  One advantage to the pattern recognition approach is that atmospheric effects would tend to wash out.  You can either make a best guess about the target (a pattern recognition capability), or you can use known spectral information to compensate for the atmosphere.  Fortunately, the Landsat Archive provides an excellent source for this information.

An important difference between Full Spectral Imaging (FSI) and Hyperspectral Imaging (HI) is that FSI deals with spectra and spectral features whereas HI is just multi-spectral imaging with lots of bands.  When dealing with bands, absolute spectral radiance is important.  When dealing with spectral features, absolute radiance is nowhere nearly as important.  This significantly simplifies instrument calibration.  As has been mentioned elsewhere, a FSI instrument is significantly easier to characterize than a multi-spectral instrument (or an HI instrument that is used as a multi-spectral instrument).

Fundamental Questions to be Answered

* How many "pseudo-invariant" target pixels will be available during the course of normal sensor operation?
* What is the variance of these pixels?

General Characteristics of an Empirical Reflectance Retrieval System

* High spatial resolution
* High spectral resolution
* Large data volume
An Application of Empirical Reflectance Retrieval is described in
Landsat for the 21st Century

DDF Proposal for Empirical Reflectance Retrieval

Related project: Measurement of Atmospheric Characteristics using Empirical Reflectance Retieval.

Related project: Real-Time Characterization and Calibration of a Full Spectral Imaging System

The concept for Empirical Reflectance Retrieval is derived from earlier work on atmospheric characterization using hyperspectral data.



This page was last modified on Wednesday, 10 November 2015