You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Mesures physiques et signatures en télédétection

application of ai techniques to infer vegetation characteristics
D.S. Kimes and J.A. Smith
NASA/Goddard Space Flight Center
Greenbelt, Maryland (USA)
Tel. 301-286-4927
P.A. Harrison
JJM Systems Inc.
1225 Jefferson Davis Hwy., Suite 412
Arlington, Virginia (USA)
P.R. Harrison
U.S. Naval Academy
Annapolis, Maryland (USA)
Traditionally, the remote sensing community has relied totally on spectral knowledge to extract vegetation
characteristics. However, there are other knowledge bases that can be used to significantly improve the accu
racy and robustness of inference techniques. Using AI (artificial intelligence) techniques a knowledge-based
system (VEG) was developed that integrates input spectral measurements with diverse knowledge bases.
These knowledge bases consist of data sets of directional reflectance measurements, knowledge from litera
ture, and knowledge from experts which are combined into an intelligent and efficient system for making
vegetation inferences. VEG accepts spectral data of an unknown target as input, determines the best tech
niques for inferring the desired vegetation characteristic(s), applies the techniques to the target dam, and pro
vides a rigorous estimate of the accuracy of the inference. Currently VEG has been developed to (1) infer
spectral hemispherical reflectance from any combination of nadir and/or off-nadir view angles (2) infer per
cent ground cover from any combination of nadir and/or off-nadir view angles (3) infer unknown view
angle(s) from known view angle(s) (known as view angle extension), and (4) discriminate between userdefined
userdefined vegetation classes using spectral and directional reflectance relationships developed from an
automated learning algorithm. The errors for these techniques were generally very good ranging between 2
and 15% (proportional rms). The system is designed to aid scientists in developing, testing, and applying new
inference techniques using directional reflectance data.
KEY WORDS: Vegetation, Directional Reflectance, Knowledge-based System, Artificial Intelligence
Few accurate inference techniques of vegetation characteristics (e.g. hemispherical reflectance, ground
cover, photosynthetic capacity, leaf area index, biomass, vegetation height, etc.) have been developed using
data that can be realistically collected from conventional sensor platforms. The success of fundamental
remote sensing science efforts that accurately infer vegetation characteristics will determine the degree and
scope of the success of science performed using Earth Observing System (EOS) data. Inference systems must
be developed that deal with the variable nature of EOS data (e.g. variable off-nadir viewing capabilities,
varying solar zenith angles, various sensor wavelengths, cloud cover problems, missing dam, and atmospheric