to 
TS. 
has 
in 
as, 
er- 
ent 
on 
)n- 
in 
1er 
)li- 
tic 
ing 
the 
to- 
gni- 
cathode-ray tube (CRT) flying-spot scanner for each 
photograph, the scan generator, and the video corre- 
lator—provides the basic terrain-surface sensing func- 
tion and produces the required error signals. The 
control computer, using these error signals, intro- 
duces model-elevation corrections which in turn cause 
the photograph motions necessary to reduce the 
sensed errors to zero. The computer also controls the 
scan generator, introducing corrections that adapt the 
scanning process to local conditions of terrain slope 
and image detail. 
The flying-spot scanners, which generate 
broadband electrical signals representing the conju- 
gate image detail, provide the basic means for auto- 
matically examining the photographs. In the analyti- 
cal stereoplotters, the optical portions of the flying- 
spot scanners are optically duplexed with the normal 
viewing system. As is shown in Figure 2, the scan 
pattern generated on the face of each CRT is opti- 
cally transferred to the photocarriage, where it is 
combined with the viewing illumination and imaged 
on the photograph, with the pattern centered on the 
optical axis. The transmitted CRT illumination is 
separated from the viewing illumination and detected 
by a photomultiplier tube (PMT). Thus, a video signal 
is produced by the PMT, while the capability for 
direct optical viewing is retained for manual opera- 
tion and monitoring. 
Visual 
Illumination 
        
  
Source 
(Yellow-Green) 
    
  
      
High Resolution 
Illumination ” 
(Violet) 
Figure 2 Flying-Spot Scanner 
The scanning process converts a differential dis- 
placement of conjugate imagery, which may be pre- 
sent in the X-direction because of an elevation error, 
to a time displacement of the video signals. Thus, in 
the processing of the video signals, relative time dis- 
placement must be measured and converted into 
signals which can be interpreted by the computer. 
This operation is performed in two parallel channels 
AUTOMATION IN COMPILATION 
within the video correlator. The first channel 1s a 
simple cross-correlator, in which the video signals are 
multiplied and time-averaged. The cross-correlator 
produces a signal which is proportional to the simi- 
larity between its input signals. A typical cross-corre- 
lation signal, plotted as a function of relative X- 
direction image displacement, is shown in Figure 3. 
Signal Output 
A 
Cross- Correlation 
  
—AX t —+AX 
  
X-Parallax Error 
Figure 3 Cross-Correlation and X-Parallax Error 
Signals for Relative X-Displacement of 
the Scanned Areas 
The cross-correlation signal provides useful in- 
formation about the degree of image match. Also 
required, however, are signals which indicate the 
direction of any matching error. These signals are 
developed in the second channel of the video corre- 
lator. The video signals are first passed through quad- 
rature networks, which phase-shift one signal 90 
degrees with respect to the other at all frequencies. 
The phase-shifted signals are then multiplied, and 
their product is resolved according to the instantane- 
ous scanning direction and averaged to produce x- 
and y-parallax error (x and y relative image-displace- 
ment error) signals. The form of the x-parallax signal 
is shown in Figure 3. For zero displacement, the 
product of the quadrature video signals is zero; how- 
ever, as time displacement is introduced by relative 
image displacement, the outputs indicate the sense 
and approximate magnitude of the error. The x- 
parallax error signal, which is related to stereomodel 
elevation error, is further analyzed to determine the 
variation of terrain elevation within the scanned area 
and produce two additional signals which define the 
X and Y components of the local terrain slope. 
One of the key features of the cross-correlation 
process that is extremely useful in the image match- 
ing application is its ability to detect signals in the 
presence of noise. Under typical conditions, the video 
signals contain substantial noise components (noise 
for this purpose is defined as any dissimilarity of the 
two video signals other than simple time displace- 
ment). Noise is introduced at the photographs by (1)