DIGITAL TECHNIQUES FOR MAP COMPILATION 
237 
an optional approach to the speed-accuracy 
compromise. 
In the predictive system, the significant 
parts of the planned system having major 
effect on time and accuracy have now been 
programmed. The description and status of 
the three program sections of this system 
follow. The use of unrectified data is planned 
and partially programmed, but it is not yet 
operational. 
System Control Program 
The system control section organizes all 
the forms of input system control data for use 
in direct contouring and orthophoto program 
sections. The data are made available in the 
form of tables and mathematical formulas, 
depending upon the most efficient manner in 
which the data can be retrieved with respect 
to required accuracy, computational speed, 
and storage requirements. 
This program functions as a central control 
monitor with respect to position or tic mark 
data and control symbols, resection-orienta 
tion parameters, ground survey data, para 
metric limit tables for use in contouring, out 
put scale of orthophoto, and non-correlatable 
areas. The tables and formulas enable opti 
mization (speed vs. accuracy) of the predic 
tive system operation when using point (as 
opposed to “area,” as in the sequential sys 
tem) correlation. 
Direct Contouring Program 
The direct contouring program has two 
parts: correlation and contouring. The predic 
tive point-correlation program initially ex 
amines the digitized left photographic data in 
an NXM area and determines the position of 
a single point within this area as possessing 
the “best” characteristic. This characteristic 
is determined by the use of one or more den 
sity (digitized transmissivity) difference func 
tions f(AD)xy, whose value is “best” in each 
NXM area in the left photo, where “best” 
implies, essentially, maximum change of 
photo density, i.e. detail. 
The function f(AD) xy at the point x, y rep 
resents a point correlation process, and the 
concept of a density function is based on the 
premise that all “x" directed changes in 
density are critical to any correlation process. 
In general, this function is simply defined as 
(fAD)xy where 
f(AD) x ,y = ADi,2 + AZ?3,4 
k 
ADu = Di — D-i and D n = H D=r+i, u +i 
2=0 1=0 
with magnitude and sign constraints on k, /, 
and A D. 
A threshold value or cutoff limit is assigned 
to the density difference function below which 
all values are rejected. This minimizes the 
necessity of accepting any system noise from 
optical, photographic, or electronic sources, 
within the photographic data. 
The subsequent f(AD) XtV to be compared 
with this initial value is at point (x"\-S x ,y) in 
the x-direction and at (x, y + S^), in the y- 
direction where S x and S y are shifts in the x 
and y direction, respectively. 
As described above, a value of the selected 
density difference of any point x, y in the 
photo relative to its neighboring points is de 
veloped from the scanned photo data. An 
optimum (empirically determined) density 
difference function is then used to examine all 
density differences and to select that point 
in NXM area which possesses the “best” 
characteristic (i.e., the one that yields the 
most valid correlations per photo). The 
sample area (NXM) is currently 36X36 
spots. 
Having searched and found a maximum of 
the appropriate density difference function of 
one point within the NXM area, the pre 
dicted parallax limits of this point must then 
be determined in order to locate the conjugate 
point in the right photograph. This deter 
mination is accomplished by first predicting 
its average elevation (using known distortions 
where required), and then predicting its 
minimum and maximum elevation, by using 
four known elevations which are adjacent to 
the point in question. Refer to Figure 12. 
These four known elevations are previously 
correlated points. Their spatial positions in 
the photograph have been determined as 
shown and lie at various points within the 
NXM areas. The “h(0)” is located some 
where within the indicated NXM area, and 
its elevation and exact position are to be deter 
mined by matching rules. 
The predicted average elevation is 
1 * 
h x ,y = — h(i). n — 4 in Figure 12 
n i= i 
The range in predicted elevation (computed 
to corresponding parallax) is then computed 
where h(n) has maximum and minimum 
values and 
where