Atmospheric effects are currently being implemented. Atmospheric techniques will make atmospheric correc 
tions to data. These techniques allow data collected by satellites and aircraft to be corrected for the atmos 
pheric effect to determine what the equivalent ground level measurements would have been. Other atmos 
pheric techniques will allow data collected at ground level to be projected to different atmospheric heights. 
Figure 1 shows a conceptual or goal decomposition of the VEG system. 
2.1. VEG as a Workbench 
The VEG system provides a workbench supporting remote sensing scientists doing analysis of optical 
reflectance data. VEG was designed using object oriented programming techniques for ease of development, 
data management and to facilitate continuing system evolution. VEG collects together in one program a 
variety of analysis techniques that were previously only available in different formats from separate 
resources. The current version of VEG has over 1500 objects including multiple rule bases, historical data 
bases, graphics tools, browsers, analysis tools, interface objects and functions. 
VEG encourages creative investigation by accommodating a variety of repeated tests under different con 
ditions. The interface acts as a research assistant in that it tells the user which combination of data is essen 
tial to a particular query, it prevents the system accepting invalid data, and it ensures that intermediate steps 
in processing data are carried out in the correct order. Each time new capabilities are added to VEG, the 
interface is expanded as well. The interface both prompts the user and validates the input. It helps the 
scientist navigate through the system reminding him or her when a step has been neglected or a value is out 
of range or of the wrong type. 
VEG manages complexity providing an appropriate level of abstraction for scientific investigation. In 
addition, calculations and data manipulations that used to take the scientist hours now takes seconds. The 
combination of complexity management and time shrinkage, enables the scientist to do more exploratory and 
"what if' thinking. It makes for more effective science. 
VEG provides a number of interesting additional features. A simple interface allows the scientist to add 
new techniques to VEG. Other interfaces allow data to be input from external files, and the results of 
analysis to be written to external files. A tool box provides capabilities to browse the system, dynamically 
plot data, get help or add text to the help system and print screen dumps. Tools are provided for managing 
the interface between VEG and the operating system so that historical databases can be used or added to dur 
ing sessions. A detailed description of VEG as a Workbench is described by Harrison et al. (1994). 
2.2. Example Session 
An example session of VEG is presented here to give a general sense of what VEG does. The inference to 
be made is view angle extension. View angle extension is inferring the reflectance(s) of unknown view 
angle(s) from the reflectances from known view angle(s). There are many remote sensing studies where view 
angle extension capabilities are required. A more detailed description is presented by Kimes et al. (1994). 
The same general scenario as presented in this example session is used by VEG to infer other vegetation 
characteristics such as percent ground cover and hemispherical reflectance. 
When the user choses goals referring to View Angle Extension, VEG presents the user with several 
options to choose from. These include (1) extension of one view angle to another view angle, (2) extension 
of multiple view angles to a another single view angle, and (3) extension of multiple view angles to another 
set of multiple view angles. Multiple view angles refer to two or more view angles and the maximum 
number of view angles is essentially any set of view angles that effectively cover the entire reflectance distri 
bution. The system then elaborates any of these top level goals into a sequence of subgoals or tasks that 
must be accomplished in order to accomplish the top level goal. 
VEG queries the user for the input directional spectral data and then characterizes the input data. For 
example, a list of assertions for one input data set might be: nadir data is available, 8 off-nadir view angles 
are available, 1 full string is in the plane 90 degrees to principal plane of sun, etc. 
Once the input data are characterized, VEG begins to characterize the target Some minimal knowledge 
about the target’s characteristics can improve the accuracy of VEG’s extended view angle(s). VEG can make 
inferences of the unknown target characteristics using a variety of techniques. For example one technique 
that it can apply uses a linear regression where the normalized difference is the independent variable and the