252 
4. A selection of DIPS II-based projects 
The following is a brief survey of research and 
development projects, which were recently 
executed on DIPS II. Further details can be found 
in the referenced literature. 
measurement by the least squares template 
matching technique (for details see Wilkins, 
1990a). Figure 6 shows a typical scanned image of 
a signalised point after digitisation. 
ULD1F - Development Lnvironmcnt I 
tmmBsammam 111 hbbibi n— 
Image xc77_cvp__1320_l: Coordinates froa Template Hatching [Pixel] 
Point no: 1B0, x = 322.326, y = 213.509, sx s B.B35, sy = 0.032 
emplate (15 x 15) 
RDI/Iwage) 
Clear] 
Cont.] 
Stat. of Fields of Ro1) 
TtZi-TTlZiT) 
Noise Analysis: 
[Grayv. Roi) (Noise Roi^ 
Non^ 
None 
Plot as Dots 
[Noibb vb. Grayvali 
Feature Type: 
HwiBument Huda: 
Location Functions: 
TM Para, for Polnta: 
Preprocessing Functions: 
Resampling Functions: 
(General HenTT][ Opt ions Menu] [Qui t^ 
H Point Q Line »»»to«-: 10^ 
O Semi Automatic 
OTemplate Hatching (Templates Menu") 
S'x Shift Siy Shift Six Scale Si y Scale Si x Shear □ y Shear 
O Radiometric Correction for Mean and Standard Deviation 
Bilinear Interpol at 1 on 
Figure 5: Large testfield for CCD camera calibration; shown within DEDIP with panels for “Interactive 
Target Location” and “Image Processing” 
4.1. CCD camera calibration 
We have two three-dimensional testfields of 
different size in use for geometric camera 
calibration. One testfield consists of a metal plate 
(size 50x50 cm 2 ) with 25 retro-reflective targets on 
it. The plate can be positioned parallel to itself so 
that a 3-D pointfield will result. The movement of 
the plate is continuous, its position is measured by 
rotary encoders. Results of calibration and point 
positioning accuracy are reported in Beyer, 
1987.Our large testfield (size 3.2 x 2 x 1.2 m 3 ), 
which consists of 136 targets marked on a wall and 
five vertical rods with 50 additional targets (all 
targets have two sizes each) each is shown in 
Figure 5. Calibration and point positioning results 
are reported in Baltsavias, Stallmann, 1990. 
4.2. High accuracy point positioning 
Two typical examples from distinctively different 
applications are reported here. One is a semi 
automatic aerial triangulation, where signalised 
points are measured in aerial photographs on the 
S9-AP by coarse operator positioning and fine 
Figure 6: Scanned image of a signalised point in 
an aerial photograph. The overlaid quad 
rangles show the location and size of the 
digitized patch before and after template 
matching. Also shown is the path which 
the template takes during the iterations 
of the matching process.