746 
NATURE 
[JUNE 2, 1923 

changes of height, either in the aeroplane or the 
ground, will cause slight changes in scale between 
successive strips, and slight persistent tilts, either in 
the camera or the ground, will cause the strips, as 
fitted in the first place, to show fictitious curvatures 
due to differences being represented toa larger scale on 
one side of the strip than on the other. 
To make a good fit between successive strips, these 
fictitious curvatures and differences of scale must, so | 
far as possible, be eliminated by distributing errors | 
between all the joins of the individual prints. We do 
this by securing to the back of the strips lengths of | 
stretched elastic, fixed to each print in one spot by 
dabs of seccotine, and, when all the strips have been so 
treated, we lay them side by side upon a table and 
stretch and bend them systematically, until we have | 

Fic. 
| 

1.—Photographic map of 7} by 15 miles, showing the elastic bands used in the compilation. 
of the distortion of the mosaics, without regard to 
scale. 
The first of these mosaics was made without the gyro 
control on the rudder ; it contains an area of ro miles 
by 5 in which no point is displaced by more than roo 
yards, but outside this area, towards the ends, there are 
points displaced by as much as 250 yards. 
In the second mosaic, which is the one illustrated in 
Fig. 1, the gyro rudder control was used, and in this 
mosaic there is no distortion greater than roo yards in 
any part. The increased accuracy is due, mainly, to 
the greater straightness of the runs, and to the pilot 
having been able to give more attention to maintaining 
height and speed constant. 
The average scale of both mosaics came out to 
1/19,800 as against 1/20,000 intended. The difference 
In the finished map the prints 
would be properly trimmed and the elastic bands removed. 
got the best general fit that can be obtained, without 
detail handling of the separate prints. 
Provided that the strips were originally taken in 
fairly straight lines, this process of systematic adjust- 
ment appears to eliminate the fictitious curvature almost 
entirely and to adjust the relative positions, even of 
points that are far apart, with remarkable accuracy. 
A final adjustment is then made, in which attention is 
given to each print separately, but no print is moved 
far from the position that it has taken up in the system- 
aticadjustment. Fig. 1 shows a mosaic laid out in this 
way. Notice the straightness of the strips due to the 
gyroscopic control, which was used in this case. 
ACCURACY OF THE Maps. 
Two mosaics of 74 by 15 miles have been compiled 
in this way, without any reference whatever to exist- 
ing maps. Some 4o points were selected on each of 
these, and their positions on the mosaic plotted on 
transparent paper. A plotting of these same points 
was then taken from the Ordnance map and enlarged 
until the best possible fit could be obtained with the 
points from the mosaic. The two plottings were then 
slid over each other until the best fit was obtained, 
and thus the remaining discrepancies give a measure 
NO. 2796, VOL. 111] 
of 1 per cent. may be due either to the aeroplane having 
been about roo feet too low, or to errors introduced 
during the systematic adjustment. 
CoMPILING TO ConTROL PorntTs. 
Our next experiment was to start again with new 
prints and to compile these two days’ work together into 
a single 15-mile. square. But this time, instead of 
fastening the ends of the elastics down to the table, we 
fastened them to laths on the edges of the table. (See 
Fig. 2.) The object of this was to enable us to apply 
systematic strains to the mosaic as a whole, after the 
first systematic adjustment, to cause it to fit control 
points. 
We chose four control points, forming a rough 10- 
mile square, and, assuming their positions to be inde- 
pendently known (in this case from the Ordnance map), 
constructed a template to fit them, upon a scale that 
i i ii i 
would most nearly fit the corresponding points upon ~ 
the mosaic, after its first adjustment. We then applied 
this template to the mosaic and found that, owing to 
the distortion of the latter, displacements of about 150 
yards were necessary at each control point to obtain an 
exact fit. These displacements were given to the mosaic 
by moving the laths as a whole, so that the adjust- 
ments were distributed over all the joins. The mosaic 
