381
the origin of
function of r.
ry. About this
(2)
(3)
gin. This with
ent angle a.
lutocollimation,
perpendicular
of images on
und is known
:h the photo-
and camera
is will need to
icient number
Equation (1)
as a research
systems. It is
nera is isolated
is found from
This is usually
ority of calibra
te practice has
leasurements on
action, aircraft
on, are looked
is an I.S.P.
papers on the
isher, 121 Half 31
the calibration
of ± 2 fx can
r, and further
is to be sure
independent
agree within
choice of glass
this to perfec-
design distor
tions down to
20 [x or less, and such lenses can for many purposes be considered as distortion
free. Using an aspherical surface on one component, a still closer approximation to
zero distortion can be obtained.
Departures from design distortion occur because of
(a) minor departures from the design values of the radii of curvature and
thicknesses of the optical elements, of the glass indices and component
separations,
(b) asymmetry in optical surfaces or the mounting of the components, wedged
filter glasses or register plates,
(c) local random variations in the optical components.
Minor departures from design values are usually kept small by the lens manu
facturer. The main effect is on the principal distance, although the distortion will
vary from lens to lens of the same nominal type. Results have been published for
measurements on numbers of lenses of the same type. Half 21 quotes results for
fifteen Aviogon cameras, which had principal distances in the range 1534 mm. to
151-9 mm.; the maximum departure of the distortion of any lens from the average
distortion was less than 10 ¡x. Hothmer 161 quotes results for other lenses, for example
Ross, Aviogon, Planigon, Metrogon, with rather similar variations of the distortion.
Asymmetry in a lens acts as though a small prism were added to a perfectly
centred lens. No single point can then be found about which the radial distortion is
completely symmetrical for all values of r and 6. A centre of symmetry, about which
the residual asymmetries are a minimum, can be found, and it can be shown 161 that
the distance between the centre of symmetry and the principal point of auto-
collimation is an adequate measure of the equivalent prism.
5. Tangential Distortion
Tangential distortion cannot arise in a perfectly symmetrical lens. For well-
made lenses, the lack of symmetry should be small; and the tangential distortion
can be deduced from the value of the equivalent prism. For a 90° lens, the maximum
tangential distortion is 04 times the separation of the centre of symmetry and the
principal point. Tangential distortion is not often measured during the calibration
process, and indeed it needs rather more elaborate equipment to do so. For this
reason alone, it is important to determine both the centre of symmetry and the
principal point during calibration and to set some limit to their separation for
acceptance of the camera.
Although the theory of the equivalent prism has often been stated, and indeed for
simple lenses it can be deduced from optical theory, there appears to have been no
attempt to justify this for the more complex lenses used in air cameras, with widely
spaced and very numerous components. Now that we have computer programmes
set up for ray tracing, it would surely be easy to calculate the effect of the de-centring
of one or more lens components. If it could be shown that each de-centring gave rise
to an equivalent prism, and that these added vectorially as do real prisms, then one
would have greater confidence that tangential distortion need not be measured during
calibration. We could also make some real effort to neutralise any residual asymmetry
in the lens, by fitting a prismatic filter of the correct angle and orientation.
6. Residual Distortions
Local variations from the design distortion can arise from non-spherical
polishing of the lens surfaces, variations of refractive index within the individual lens
components, local departures of flatness in the focal plane, and so on. When
calibrating a camera, the measurements also show random deviations arising from