ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002
Due to these two disadvantages, the use of non-metric
cameras in photogrammetric applications is considered under
very strict constraints. However if non-metric cameras are
calibrated properly they could be used in more applications.
In section 3.1 a brief summary of the systematic errors
introduced by the non-metric cameras is presented. Section
3.2 describes the mathematical models employed to
transform image space coordinates to the object space
coordinates taking into account the effect of the systematic
errors.
3.1 Systematic Errors Introduced By Non-Metric
Cameras
"Systematic errors in non-metric cameras are the deviation of
the physical imaging event from the projection mathematical
model", Karara (1989). There are two types of systematic
errors; the external systematic errors, and the internal
systematic errors. The sources of the external systematic
errors are: the atmospheric refraction and the comparator
errors. However, according to (Marzan and Karara, 1976) the
atmospheric refraction errors are neglected for object
distances less than 300 meters. The causes of the internal
systematic errors are: the lens distortion, the film
deformation, and film unflatness.
Lens distortion is considered as the major source of errors for
the non-metric cameras. Lens distortion is introduced by two
components: symmetric lens distortion and asymmetric lens
distortion. The adopted mathematical model for symmetric
lens distortion according to (Brown, 1971) is given by
Equation (5).
Ar = kyr + kor +r” +... (5)
ere k „‚k „‚k , 7 constants,
Wieie t 25.
r ^ radial distance from the point of symmetry, and
Ar = symmetric lens correction in the radial direction
The asymmetrical lens distortion is due to the lenses
decentering. We employed the correction model presented by
(Faig and Moniwa, 1972), Equations (6).
Ax = by (r? +2x2)+2byxy
(6)
Ay = by (r? +2y°)+2b,xy
Where b D b, — constants,
r — radial distance from the point of symmetry,
x, y 7 point coordinates in image space, and
Ax, Ay = symmetric lens corrections
The film deformation inside the camera is caused by bad
flatness of the film. The film deformation outside the camera
is due to lack of control over temperature, and the humidity
during processing and storage of non-metric films. In
(Robson, 1992) the 6-paramer transformation model is used
to model the total film deformation.
3.2 Transformation Between Image Coordinates and
Ground Coordinates
We employed four models to transform the image
coordinates and the object coordinates. The first model is the
collinearity representation with no systematic error
correction parameters. Nine transformation parameters are
included in this model. The second model is a modified
version of the collinearity representation, 6 more parameters
are added to handle the systematic errors, Equation (7). The
third model is the DLT representation, (Abdel-Aziz and
Karara, 1971). The fourth model is an adapted form of the
DLT representation, Equation (8). The DLT is a linear
treatment of a non-linear problem so its results are
approximate.
a, *axa,yt kar? -
my =X) mo(Y -Y.)* ma -Z.)
m3yX —X,)* myyY -Y,) t ma33Z —-Z.)
b, bx by ko yr! -
aX - X) myyY -Y.)* mayZ 7Z«)
my (X — X) *myyY - Y.) + M33(Z = Ze)
(7)
Where f — camera interior parameters,
x. Y, Zz, = exposure station coordinates,
x, y = image space coordinates,
X, Y, Z = object space coordinates,
m through m. rotation matrix elements,
r = radial distance from the point of symmetry, and
k,k,b,b,b ,a ,a,a =systematic error
12 ro iy ela
correction parameters
xlX LY +I3Z la loxX * lgxY el XZ + xr? 0
V+ Is X +16Y +17Z + Ig +1gxX + [i xY ++I xZ ++kpyr? = 0
(8)
Where x, y ^ image space coordinates,
X, Y, Z ^ object space coordinates,
I through un = DLT parameters,
r = radial distance from the point of symmetry, and
k, k, = systematic error correction parameters
4. DIGITAL APPROACH FOR OBJECT SPACE
RECONS TRUCTION
Two object space reconstruction systems are presented in this
paper. The first system employs a cheap 35mm camera, Zenit
E, and an inexpensive flatbed scanner, while the second
system uses only a low-resolution digital camera, Casio QV-
A10. First 50 target points are fixed as shown in Figure 3.
The points are fixed on wood plates with different thickness
varying from 5 mm to 10cm. A theodolite and a total station
are employed to find the reference ground coordinates of the
target points.