Modeling the performances
infrared scanners and cameras in POVRAY
Only since about 2003 has it
possible to make thermal cameras with 'staring' focal planes.
Prior to that,
it was necessary to scan a
very small number of detector
elements through the field of view with either a
or a 2-dimensional scanner.
This is a model of a
1980s 2-dimensional infrared scanner which scanned a 23
linear detector through a 600,000 pixel field. The hexagon
9,333 rpm and the pair of yoke mirrors at 25Hz
This shows the type
which early ( 1980s) thermal imagers made. The camera is
on an anti-tank missile control post and present the image to
gunner who is controlling the missile attacking the tank
The tank detects the approaching missile and deploys smoke and an
movie with an
ideal camera and then added off-line the defects created by the
scanning geometry and the detector non-uniformity and temporal noise.
As detectors improved
only 1-D scanning was required. Such a camera was
the POVRAY scenes page. Radiometric errors were still present and
appear as the
The ultimate performance of
modern thermal imaging cameras is generally determined by the MTF of
the complete system and the random noise from the detector.
'Background Limited' camera, the noise is set by the noise in the
photons rate from the scene. Generally the output is observed
human eye and brain which acts as a matched filter to the image
modulation and the noise. The performance as a device capable of
imaging temperature modulation in the scene is characterized by the
'Minimum Resolveable Temperature Difference' as a function of the
temperature contrast in a 7-bar
chart object in the scene.
movies I've modelled a complete thermal camera with
in the scene and post-processed to apply noise and MTF. Notice how the
visibility of the bar charts increases with the moving image. These
used to predict the performance of the real camera.
agreement with the real camera was obtained.
The ability to visualise stray
light is one of POVRAY's greatest features. Stray light is
critical in the assessment of infrared lenses. It manifests
itself in two ways. Firstly as false images created by bright
objects in the scene, and secondly by the radiation from the lenses and
metalwork reaching the focal surface by either/both
refraction/reflection/diffraction in and from the lenses and
metalwork. This latter is known as the Narcissus
Generally a cold spot appears in the centre of the image.
removed by 'Non-Uniformity correction' algorithms, but if the scene or
camera temperature changes the correction balance is
Bear in mind that a thermal camera is required to detect contrasts
typically less than 0.5% and is sensitive to contrast of a few
hundredths of a Kelvin at a distance of several kilometres - even
miniscule changes in temperature of
the scene or the camera can become visible as image defects. These defects
are well known
but difficult to model, although there are many programs ( I've used
Nathan Kopp's POVRAY version in Ken Colefax's 'cities' model
for the cosmetic cases when 'cheating' is allowed.
at Straylight 2.6MB
proper scientific analysis, photon mapping is essential.
This is an
example. I used the awesome routines
created by Jaime
Vives Piqueres at
to create the 'optics lab' from my desk position in the office to
contain the optical bench and electronics.
shows an 8-element F/1 lens with a 150degree field of view on
The lens has three
aspherical surfaces and one diffractive. The lens surfaces have angle
dependent reflectivity according to the designs of the anti-reflection
coatings. The cryogenically cooled
focal plane is shown glowing 'blue'. The interior of the lens is shown
1.3MBThis following movie is a
photon-mapping rendering of the image formed by the lens at the
diagonal of a square focal plane as an object moves through the
150degree field. This again confirms the validity of the
mapping algorithm implementation algorithm in
notice that the image is not a collection of 'spots' in the spotty
diagrams typical of traditional ray-tracing programs - it has the
appearance of a real image i.e of a limitless number of spots.
'assumed gamma' setting and using a heightfield tga image it
possible to create a dynamic range of 10^7. I then read the
files with the 'jet' palette to cover that range. Notice how
very wide dynamic range has increased the visibility of the aberrations
increasing with field angle.
little difference between these images and those calculated in the
top-of-the-range optics modeling program ASAP.
This shows a ZEMAX
of the causes of the external stray light images. A point light source
to the left is collimated by a lens. The test lens rotates around the
entrance pupil created by a crygenic stop just prior to the focal
plane. As the lens rotates a myriad of unintended paths are
formed by reflections at imperfect 'anti-reflection' coatings and total
internal reflection inside the lens elements.
This is a model of one section
submarine periscope known to have a weak but still noticeable
straylight problem. The layout includes a number of lenses
prisms. We believed that the straylight was caused by a
Dove prism used to steer the sightline in elevation.
movie, a screen plane moves
through the top components to arrive at the first focused
intermediate image surface. There are five 'suns' in the
As the suns move
the field of view the effects of light following unanticipated paths
creates multiple images of the suns.
and as the suns
move in elevation and the steering prism rotates, more straylight
Glint from lenses
can also be
important. In this scene a lens is illuminated by an axial
and light is reflected from the lenses.