23 
S, 
5 
be 
r- 
els 
nner 
ter- 
A 
pro- 
reas 
tion 
ably 
the 
ang- 
and 
in 
ter 
cu r- 
in- 
port 
the 
9524 
n an 
at 
data 
with 
1 in 
juid- 
ygon 
jery. 
ion, 
)ara- 
3 nu- 
nents 
r ip- 
sys- 
and 
de- 
d in 
base 
base 
iting 
t of 
field 
The objective of this last stage is to select, prioritize 
and display the set of discrete, raultispectrally separable 
targets for narrow-band data collection. Following the man 
ual encoding of target boundaries through direct stereo di 
gitization, descriptive characteristics of each target with 
in each spectrally independent polygon are visually inter 
preted and encoded. These descriptors are keyed into the 
data base and referenced to the previously digitized geo 
graphical base file. A positionally accurate digital data 
base of statistically representative, spectrally homogeneous 
target data at user determined scales and map projections is 
in this way developed. 
It is then possible to produce a plotter-drawn overlay of 
targets with desired attributes. The synchronously flown 
CIR photography could be used at an enlarged scale of 1:1000 
as base imagery, upon which the outlines of individual spec 
tral targets can be overlain. These are labeled sequential 
ly to guide data collection by charting the course of the 
mobile spectral radiometer. In this way, a data base capa 
ble of indexing sequentially ordered spectral data is estab 
lished, at the same time an empirically derived representa 
tive sampling frame is produced. The geocoded data base and 
associated query algorithm permit the development of summary 
sampling statistics in tabular, graphic or cartographic 
form. 
The groundwork is laid for the visitation of as many targets 
for spectral measurements as is necessary to acquire a rep 
resentative sample for subsequent sensor band width charac 
terization. This is planned as a user-defined determination 
to be handled interactively by the system. The analyst spe 
cifies either the number of sites to be visited per 15' 
quadrangle or the percent of areal coverage desired, and the 
system responds with a user-specified selection of data col 
lection support materials. This could be as brief as a 
printout of targets, their locational coordinates and a 
short description of each, or as elaborate as a complete 
sampling itenerary replete with an indexed routing map, en 
larged photo coverage overlayed with indexed plots of sta 
tistically selected target areas, and a tabular report of 
the sample selected, including site descriptors, areages, 
and associated statistics. At this point, spectral data 
collection can begin with the assurance that time in the 
field will be used efficiently, and that data collected will 
be representative. 
With the criteria of spectral and positional accuracy upheld 
throughout the design, it is possible to extend, with confi 
dence, the results of the spectral sampling to the entire 
study area. Of course, only a test of these recommendations 
will satisfactorily determine their accuracy and validity 
under the fire of an operational research program. It is 
precisely this that is planned as the next step in the sen 
sor characterization research program at NASA-Goddard Space 
Flight Center.