vely 
f 10 
ning 
ters 
> ap- 
lane 
f the 
LG. 
or is 
able 
ents 
sta- 
Il as 
nsor 
aces 
> Te- 
ner- 
tral 
seo- 
for- 
the 
| are 
) re- 
, the 
‘the 
it of 
the 
ther 
the 
COI- 
s to 
ime 
> at- 
ner- 
the 
y to 
atly 
2 Illumination 
Object 
Optics 
Sensor 
Camera electronics 
  
Signal transmission 
Frame grabber 
  
  
Digital image 
Target location 
Positioning 
  
  
  
  
Figure 1 Image acquisition with solid-state imaging 
Sensors. 
2 CHARACTERISTICS OF 
COMPONENTS 
2.1 Illumination 
The importance of the illumination is generally underes- 
timated. The temporal stability, the spectral characteris- 
tics, and the distribution of the light intensity on the 
object affect the measurement accuracy in several ways. 
The temporal stability is critical for very short, i.e. less 
than the power supply frequency, and very long image 
acquisition periods. Shuttered cameras (mechanical or 
electronic) need to be synchronized to the power supply 
of the lighting system when illumination sources which 
vary their light intensity with the frequency of the power 
supply are used. Electronically triggered fluorescent 
lights limit these variations to a few percent. Imaging 
systems requiring longer time spans for image acquisi- 
tion, such as cameras with micro-displacement of the 
sensor (e.g. ProgRes 3000) or cameras with area and/or 
line sensors scanning a larger focal plane require an ex- 
cellent long-term stability of the illumination for time 
spans ranging from several seconds to thirty minutes. 
The precise effect of the variation in illumination intensi- 
ty on the accuracy depends on the type of sensor and the 
target location algorithm. When employing methods 
which are susceptible to illumination gradients across the 
area of features, e.g. centroiding or Least Squares Match- 
ing (LSM, Gruen, 1985), such variations must not induce 
illumination intensity gradients across the features. Ad- 
joining pixels in imagery acquired with micro- 
positioning cameras will exhibit the variations over the 
complete acquisition time. Adjoining pixels of images 
acquired with line-scan cameras exhibit the variations 
occurring between the acquisition of individual lines. 
The spectral characteristics of the illumination affect the 
photo response non-uniformity (PRNU) of the sensor as 
well as the modulation transfer function (MTF). The ef- 
fect on the PRNU is considered to be negligible at this 
point as the PRNU of current solid-state sensors (pro- 
duced for broadcasting applications) is better than 1% 
(see following discussion of PRNU). Optical crosstalk 
leads to a degradation of the MTF at longer wavelengths 
for front illuminated solid-state sensors. Many cameras 
do thus use an infrared (IR) cut filter to eliminate light 
with a wavelength longer than 800 nm, which has a ab- 
sorption length (distance where 50% of photons were ab- 
sorbed) of 10 pm in silicon. 
The most obvious problem associated with the illumina- 
tion is the variation of its intensity due to inherent prop- 
erties of the illumination and/or shadows. Only the 
gradient induced across a feature of interest is of impor- 
tance (Global differences do not lead to positional chang- 
es with typical target location algorithms like Least 
Squares Matching). Thus shadows, i.e. borders of shad- 
ows, are of great importance. Latter is dependent on the 
type of targets and illumination to be used. Shadows are 
a significant problem with standard targets but virtually 
nonexistent with retroreflective targets, which can in turn 
be affected by local variations in their reflectivity thus in- 
ducing similar problems (e.g. oily substances on retro 
targets). The different reflectivity of the surrounding ar- 
eas of objects has also been found to lead to gradients of 
the illumination intensity on the object which can result 
in displacements of several hundreds of a pixel (Beyer, 
1992a) 
2.2 Object / Targets 
The type, color, size, and form of a target do obviously 
affect the accuracy. Type (material) and color are factors 
defining the contrast of the target with respect to the 
background. Retroreflective targets exhibit the strongest 
return and can thus be used to provide the highest con- 
trast. They can under certain conditions also exhibit more 
uniform reflective characteristics across the field of view 
as the degradation of the light intensity to the image cor- 
ners due to the optical system can be counteracted via the 
stronger response as the angle between the illumination 
and the imaging rays becomes more optimal to the sides 
of the FOV (field of view), resulting in a stronger return. 
They are furthermore illuminated by lights on or close to 
the optical path of the camera, thus eliminating shadows. 
The target size must be considered together with imaging 
scale. Figure 2 gives an empirically determined relation 
between the target size and the internal precision of tar- 
get location. Figure 2 a shows the targets ranging in di- 
ameter from 2 to 17 pixel. The diameter increases by a 
factor of y^ from target to target. The plot of the inner 
precision indicates the strong improvement for target di- 
ameters from 2 to 6 pixels. An internal precision of 0.005