■■■■I
SV CYCLOP: A New Instrument for Close-Range Photogrammetry
743
the plotting reference system and a reduction of the base with respect to the real displacements of the two projection centres 5.
If (p is lower of 5 gon, considering the base to be equal to the displacement 5, the (p rotation causes systematic errors that are lower
than the m.s.e. of the calibrated digital cameras. If this hypothesis cannot be accepted, the knowledge of a distance on the object al
lows the correct scaling of the stereomodel. CYCLOP assures that the acquired images have the same cp rotation: dcpl and d (p 2 in
equation (1) are null as is the y-parallaxes.
The two images acquired using CYCLOP usually both have the previously described elementary rotation but the resulting stereo
model is free from the y-parallax.
Bar of CYCLOP
Camera axe
Fig. 4: Inclination of the camera axes around X
3. EXPERIMENTAL TESTS
Fig. 5 Inclination of the camera axes around Y
In order to verify the stability of CYCLOP, a series of takings was performed using a NIKON D1 camera equipped with a
NIKKOR/28 mm lens; the camera-lens system was calibrated (principal distance = 28,981 mm; radial distortion curve coefficients
ki= -1,52694-10' 8 -k 2 = 3,45676-1 O' 15 ).
Two different calibrated sets of points were used for the tests. The first consisted of 30 points regularly spaced over a 15 m side cube
that simulated a standard building (see figure 6).
The second set of points consisted of a steel plate (20 cm x 20 cm) with an engraved regular grid of points spaced at 1 cm; some tar
gets of different calibrated heights permitted the simulation of the survey of small objects (see figure 7).
Fig. 6: Test field simulating the survey of a building
Fig. 7: Test field simulating the survey of
small objects
The first performed test was that of the computation of a relative orientation of the stereopair acquired using CYCLOP. In all cases
the angular differences was less than 0.01 gon and the residual parallaxes were less than 0.3 pixel.
Table 1 shows the results for one of the analysed stereopair.