829 
otography 
comparable 
s that the 
etation 
tance 
ical- 
aphy of an 
ical- 
PROBABILISTIC IDENTIFICATION KEYS 
William H. Anderson 
Technicolor Graphic Services, Inc. 
EROS Data Center 
ABSTRACT 
and equip- 
and Wild- 
by Surdex 
This paper describes the logic behind the construction and use of 
probabilistic airphoto identification keys. A hypothetical example is 
presented which demonstrates the basic statistical framework of their 
construction. 
Bavesian probability theory is used to modify a priori information 
according to the results of densitometric measurements taken from the 
aerial photographs. This approach is shown to be useful when operating 
with minimal-level diagnostic differences between candidate identifications. 
ardwood 
ngs of 
eting, 
INTRODUCTION 
The concept of using a key as an aid to identifying objects and 
Publica- 
conditions originated long before the era of aerial photography. Voss 
(1952) in his excellent paper discussing the history of keys in systematic 
biology, attributes Aristotle with taking the first major steps towards a 
systematized classification of natural objects. Through the efforts of a 
lication 
long and distinguished line of systematic biologists, an efficient and 
effective set of procedures has evolved which can be used to guide the 
cel 
construction of identification keys. These concepts are applicable to 
the development of keys in any discipline, including photographic 
interpretation. 
hods o 
Many documents exist today where pertinent descriptive air photo 
information is variously organized and arranged. Such documents are 
943. 
cing of 
produced in order to present to the reader (user) a systematic and, 
hopefully, useful presentation of information that can be used to identify 
objects found on the aerial image. Most aids qualify as "keys" although 
vast differences exist in respective usefulnesses. This paper focuses on 
a type of key incorporating candidate object probabilities in its design. 
An argument is presented supporting the hypothesis that probabilistic keys 
photo- 
are the most useful and accurate of all air photo identification key types 
under certain common situations. 
. John 
Experience has shown that it is best to tailor keys individually 
for each local area. A typical probabilistic key framework is demonstrated 
in order to illustrate the manner in which pertinent local information is 
used in their construction. It will be shown that the probabilistic key 
approach is strongly oriented towards the production of keys that indeed 
maximize the use of locally valid information.