places the operator will contain a complex array of 
sensors and decision elements. 
The automation problem is further complicated by 
the nature of the input information, the image detail 
on the aerial photographs. The principal difficulty 
here is with the natural scene rather than with the 
recording system. Modern aerial-camera systems and 
photographic films are capable of recording an im- 
mense amount of image detail—-on the order of 10% 
elements per square millimeter. However, the quan- 
tity and characteristics of the information available 
for recording vary greatly within an average scene. As 
a result, the photographs typically contain a number 
of areas that are difficult to interpret because the 
image detail exhibits low contrast or high density. To 
maintain an established level of performance, the 
instrument operator must adapt to these situations, 
perhaps by varying his plotting speed, or by interpo- 
lating over larger or smaller image areas, or simply by 
changing his level of concentration on the stereoper- 
ception task. To achieve a reasonable level of inde- 
pendence from the characteristics of the photograph, 
the automated system must possess similar adaptive 
capabilities. 
In view of the variety and complexity of the tasks 
performed by the instrument operator, it is perhaps 
not surprising that no effective means have been de- 
veloped for completely automating his functions. 
However, substantial progress has been made with 
systems which provide partial automation by relieving 
the operator of the more routine and mechanical 
tasks. A major advance in this area was made in 1958 
by G. L. Hobrough, then of Hunting Associates 
Limited, Toronto, Canada, who successfully applied 
electronic scanning and correlation techniques to 
provide automatic stereoperception for a projection 
stereoplotter. Initial demonstrations of automatic 
profiling and contouring were also performed by 
Hobrough, using analog computing and control equip- 
ment in conjunction with the plotter. 12 
Starting in 1963, an automatic plotting system for 
the analytical stereoplotters was developed at Bendix 
Research Laboratories. In this system, described in 
the body of this paper, a high degree of automation 
has been achieved for the stereoperception and plot- 
ting functions, using electronic scanning and correla- 
tion techniques similar to those developed by 
Hobrough and expanding the functions of the control 
computer. The availability of the stored-program 
control computer has also enabled significant ad- 
vances in automating overall control functions and in 
providing the system adaptivity required to handle 
difficult photographic and terrain conditions. 
Several automated analytical stereoplotters are 
currently in service. In most situations, instruments 
have demonstrated a level of accuracy comparable to 
that of the manually operated analytical plotters. 
However, experience with these instruments has 
pointed to several areas of potential improvement in 
automatic stereoperception and plotting. These areas, 
along with several possibilities for expanding the over- 
all system capabilities, are discussed in a subsequent 
section of this paper. 
Automation of the photointerpretation function 
by pattern-recognition techniques, an area of con- 
siderable recent research activity, is not discussed in 
detail in this paper, since it is the subject of another 
paper in this volume.* However, the potential appli- 
cations for these techniques in relation to automatic 
plotting are noted in the conclusions. 
AUTOMATIC STEREOPERCEPTION 
In performing stereoperception, the operator 
senses any deviation in the position of the reference 
mark with respect to the apparent terrain surface and 
introduces the corrections required to place the mark 
in contact with the surface. Referred to the photo- 
graphs, this operation implies that conjugate imagery 
on the two photographs is matched (that is, adjusted 
to remove any relative X-Y displacement) at the 
optical axis which corresponds to the reference mark. 
Thus, a system which performs automatic stereoper- 
ception must be capable of examining the imagery on 
each photograph at or near the optical axis, generat- 
ing error signals indicating any matching errors, and 
introducing the corrections required to remove these 
errors. These operations, of course, must be inde- 
pendent of which point in the stereomodel is being 
examined, or what type of image detail happens to be 
present on the photographs. 
A block diagram of the system which pertorms the 
automatic stereoperception function for the auto- 
mated analytical stereoplotters is shown in Figure 1. 
The image correlation system—which consists of a 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Scan Video 
Scan Signals Flying-Spot Signals Video 
Generator Scanners Correlator 
Viewer 
: Cross- Correlation, 
Sen Se Parallax Errors, 
an ape Photograph Slope Errors 
Motions 
Control 
Computer 
  
  
  
Figure 1 Automatic Stereoperception System— 
Functional Organization 
*R. M. Centner and E. D. Hietanen, Automatic Pattern Recogni- 
tion Techniques for Photographic Analysis." 
CHAPELLE, WHITESIDE, AND BYBEE 
cati 
phc 
latc 
ti01 
con 
duc 
the 
sen 
sca 
Scal 
anc 
bro 
gat 
cal 
spo 
vie 
pat 
call 
con 
on 
opt 
sep 
by 
is 
dire 
tio 
  
Al