f the
and
ation
heap
nited
was
cost
n of
with
ther
was
and
itrol
] by
it in
inda
]life
this
the
the
tion
ried
1bia
the
ield
ital
the
ster
the
ion.
will
trol
has
applied one of the first graphical methods of strip
triangulation, called hand templet method [Manual of
Photogrammetry, 1980]. In this method, each photograph
is replaced by a transparent templet on which lines have
been drawn from the principal point radially through each
image point (field control and tie points - both natural
features). The templets are assembled and the intersection
of all radials common to each identified image point
represents the adjusted planimetric position of each point.
The points of radial intersection are transferred to the base
sheet by pricking through the templets. The whole set of
points is determined in an unknown but uniform scale.
Therefore, they have to be subsequently transformed to the
ground coordinate system through the use of field control.
In the original method, which was used more than fifty
years ago, both steps concerning assembly of templets and
transformation of determined control points to the field
coordinate system had been graphically done. In the
proposed modified version, the second step is executed with
a computer programe of similarity transformation. The
planimetric field control can be determined with the use of
low accuracy GPS system. The number and distribution of
field control for block aerial triangulation based on the
above approach, has to satisfied the requirements
concerning field control for triangulation of separate
strips.
Mapping of features. Vegetation boundaries and other
required features, which do not exist on 1:50000
topographic map, can be mapped from aerial photographs
with the use of simple and cheap optical device known as a
Sketchmaster. The planimetric positions of at least four
photo control points for each photograph, determined by
graphical-analytical triangulation, should be used to rectify
each of photograph for tilt and scale corrections. After
orientation of the photo, the photo control images match
with their corresponding field points, marked at map scale.
Subsequently, all new features are transferred from
photograph to a map overlay.
The above approaches for control densification and
mapping of features can be used only for terrain with
moderate relief, since the relief displacements are neglected
in compilation. Assuming that all precautions are taken in
the discussed graphical strip triangulation and mapping
procedure, the final accuracy of triangulated photo points
and mapped features would not exceed the graphical
accuracy of points/features at scale of used photographs.
Therefore, to obtain the acceptable accuracy of
compilation, the scale of photographs should be larger or
at least the same than the map scale.
3. DESCRIPTION OF EXPERIMENT
The area of the Kasanka National Park and surrounding
Kafinda Game Management Area in Zambia has size of
29.5 km in North-South and 42.5 km in East-West. The
relief of this terrain is moderate with 100 meters
differences in elevation around the area. Four sheets
of existing map at 1:50000 scale for this area were
compiled in 1967 and since then have not been updated.
Therefore, for creation of a GIS data base for this area,
there was need to determine all missing features on the
map, such as new roads and settlements as well as
155
boundaries of all vegetation polygons. For acquisition of
these data, simple photogrammetric approaches presented
in chapter 2, were applied.
A block of 86 aerial photographs distributed in six strips,
at 1:30000 scale, taken in May 1990, with 60 % end lap
and 30 % side lap, were used. For purpose of this project,
43 natural field control points had been identified in field
and on photographs around the Park area, and their
planimetric positions were surveyed with a low accuracy
Trimble Ensign Global Positioning receiver. The X and Y
accuracy of GPS control points, estimated from multi
measurement of ten known ground points, were 14 and 16
meters, respectively. This corresponds to 0.27 and 0.31
milimeters at 1:50000 map. For densification of GPS
planimetric field control to the total number of 170 tie
points - photo control (natural features), distributed
uniformly around all photographs, the modified hand
templet aerial triangulation was applied. As a result, for
each involved photograph, six planimetric photo control
points were determined. The accuracy of the applied
method was estimated from residuals obtained for 43 GPS
field control points, after analytical transformation of each
triangulated strip to the ground coordinate system. The
average accuracy for these points distributed in six strips
was 15 and 13 meters for X and Y coordinates,
respectively. This corresponds to 0.30 and 0.26 milimeters
at 1:50000 map and 0.5 and 0.4 at photo scale 1:30000.
The received accuracy of triangulation for GPS points was
within the same range as the accuracy of their field GPS
determination. Similar accuracy can be expected for tie
points (photo control) since both GPS ground control and
triangulated photo control were natural features’ points of
similar quality. For tie points, common for each two strips,
the mean square errors were also calculated from the
differences in their positions obtained after transformation
of each strip to the ground coordinate system. The avreage
mean square errors for all tie points of a block are, for X
and Y 17 and 14 meters (0.34 and 0.28 mm at map scale),
respectively [Bujakiewicz, 1993]. The positions of GPS
ground control and triangulated photo control points
within the Park area, covered by photographs distributed
in six strips, are shown in figure 1.
Another experiment was carried out in addition in the
Department, to compare the accuracy of the modified
graphical triangulation with conventional independent
models method [Musonda, 1994]. One strip of 13
photographs at 1:20000 scale of East part of Zambia was
triangulated twice with the above methods. For
independent models triangulation, a computer supported
stereoplotter A8 was used for models formation and PAT
MR software for their simultaneous adjustment. Results
from the modified graphical triangulation were compared
with those from independent models method (assumed as
accurate). The mean square errors estimated from
differences between the corresponding values for X and Y
coordinates of 40 tie points, were 7.1 and 7.4 meters (0.15
milimeters for X and Y at 1:50000 map scale or 0.35
milimeters for X and Y at scale of used photographs). Ten
field control points were field premarked and determined
with field methods. In this experiment, the tie points,
triangulated with the use of templet method, were
determined with higher accuracy because of their better
quality and the fact that the field control points were
premarked and surveyed in field with higher accuracy.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
