, Was 
The 
With 
ction, 
ratio 
ht of 
ound 
phy’, 
work. 
96). 
I 
co 
ae 
co 
| 
AB 
© 
| 
-— 
N 
— 
| 
C1 
rs 
| 
Cc 
— 
FLL 
— 
e 
  
| 
  
  
  
  
  
    
  
  
  
  
Raskelf, North Yorkshire 
| 900,000 m 1000.000 m 
E = 
S S 
3 S 
= 3 
900.000 m | d onmia dudas 1000.000.m 
Scale 1: 1250 
PE RR 
20m 0 20 40 60 80m 
  
  
Figure 3: Orthophotomap with vector overlay of Raskelf. 
4.4 Survey Accuracy 
The accuracy and repeatability results for the 
stereoscopic and orthophoto measurements can be seen 
in Table 4. Strangely the repeatability of the 
measurements was slightly worse than for the Newcastle 
survey. Although these tests were not exhaustive, the 
greater clarity provided by the larger scale imagery was 
expected to provide more reproducible measurements. 
Nevertheless they are encouraging as they compare 
directly to those reported by Fraser and Shortis (1995) 
using a monochrome DCS200 camera which 
theoretically should provide a higher resolution than its 
colour counterpart. 
The stereoscopic accuracy figures reflect the 
effectiveness of the zoom facility. The theoretical 
accuracy of the DMS as stated earlier should provide a 
heighting accuracy of 0.64 m (1 pixel / base to height 
ratio). This has been improved to 0.25 m by measuring 
parallax to the sub-pixel level using the zoom tool (0.25 
pixel at 4 times magnification). The relatively poor 
performance in terms of external accuracy for heighting 
from the DEM can be attributed to the post spacing and 
subsequent interpolation of heights from this. 
The fact that the orthophoto measurements are less 
accurate than the stereoscopic measurements can again 
567 
be attributed to the DEM post spacing and subsequent 
interpolation. The error in x is noticeably worse than the 
error in y for the measurements made from the 
orthophoto. This could be related to the heighting errors 
in the DEM, causing the failure to totally remove x- 
parallax from the orthophoto. 
All of the errors quoted for the mapping come from data 
that has been observed with the highest measuring care. 
For an absolute Quality Assurance (QA) check on the 
planimetric accuracy of the mapping produced in Figure 
3, the ground survey measurements were compared with 
the vector detail after editing in AutoCAD. This is an 
absolute check on the entire mapping flowline. The 
check (on 21 common points) yielded an RMS error of 
0.5 m in plan. This is just outside the OS quoted error of 
0.4 m for their Land Line digital data. Reasons for the 
degradation of the planimetric accuracy (from 0.2 m to 
0.5 m) can be attributed to the editing of the data 
(squaring buildings, extending lines etc.) and the fact that 
some points may not be exactly those that were surveyed 
(for example when plotting a wall, the centre was plotted 
as opposed to the outside corner which was surveyed). 
In addition to this, zooming in to a magnification of four 
times is impractical in many cases since only a small 
area is covered on the screen at any one time. As a 
result, much of the mapping was carried out under two 
times zoom which may have degraded the accuracy. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996