highly reliable range measurements are necessary for geome- 
tric inspection of tunnel tubes. 
Gap-less inspection of tunnel tubes requires a range of up to 
10 m, a spatial resolution of 2500 pixels per 360° profile, and 
a distance between two consecutive measured profiles of less 
than 2.5 cm. Due to sooty walls and metallic objects in the 
tunnel, the sensor system has to handle high dynamic range re- 
flectances of the objects. Furthermore, the sensor system must 
be robust when dealing with environmental influences such as 
temperature or humidity as well as varying illumination 
conditions (dark, ambient light, lamps, etc) as they are typical 
for tunnel tubes. Safe operation with respect to people is re- 
quired all times. 
Only non-tactile sensors are suited for covering these de- 
mands. Non-tactile range measurement techniques may be 
classified as either active techniques, directing visible or infra- 
red (IR) light, ultrasonic [16] or radar [17] pulses to the sur- 
face to be measured, or as passive techniques based on vision. 
A rich variety of passive vision techniques produce three- 
dimensional information. Traditionally they have lacked ro- 
bustness with changing illumination conditions and generality, 
and have not proven themselves effective in practice. Passive 
stereo or motion stereo vision [18,19] are particullary pro- 
mising sources of range information, but require substantial 
data processing to match images with each other in order to 
determine range by triangulation and, therefore, are not well 
suited for real-time tunnel surface inspection. 
For these reasons we have selected an active range measure- 
ment technique which directly determines "range" data with a 
minimum computation time. To achieve high spatial réso- 
lution, only an active technique emitting collimated laser light 
is suitable. The collimated laser beam is directed to the target 
to be measured and the back scattered light is sensed. In ad- 
dition to "range" measurement, evaluation of the magnitude of 
the back-scattered light provides an "active grey level" which 
is similar to the grey level information of a video camera. 
Both range and grey level data of a target point are registered 
at the same time and correspond to a single target point 
defined by the laser beam direction. Due to emitted laser 
energy, both informations, range, and grey level data are near- 
ly independent from environmental influences and illumina- 
tion conditions. 
In order to achieve profile data the range measuring system is 
combined with a one-dimensional scanner for 360? beam 
deflection. For longitudinal inspection of the tube, the system 
is mounted on a special vehicle (or train) navigating at a maxi- 
mum speed of 5 m/s through the tunnel. Resulting spiral pro- 
files are combined with respect to the corresponding sensor 
positions. The final range image reflects geometric dimensions 
of the tunnel tube whereas the grey level image is used for vi- 
sual inspection, surface classification, and documentation pur- 
poses. 
2.2 The two-frequency phase-shift 
method 
Because of the requirements for high-precision range mea- 
surement within a range of up to 10 meters, evaluation of the 
phase-shift (Fig. 1) between a reference laser beam and the 
back scattered laser light is more suitable than measuring the 
light's extremely short time of flight. 
The amplitude P, of the emitted continuous-wave laser signal 
is simultaneously intensity modulated (am-cw) with two fre- 
quencies Q, and o». Laser light back scattered from a target is 
collected by an avalanche photodiode. The amplitudes Py are 
fairly small. Due to the time-of-flight, the received sine- 
shaped signals are phase-shifted in relation to their reference 
in the transmitted signal. Phase shifts ¢, and ¢, are propor- 
tional to the range d and the modulation frequencies. Since 
phase shifts are only unique modulo 2m, the modulation fre- 
quencies ®; and ®, are selected to provide a sufficient mea- 
surement range with an appropriate range resolution. A low- 
frequency signal (LFS: o, = 10 MHz) guarantees a coarse but 
absolute measurement range of s, (s, 2 15 m), whereas the 
high-frequency component (HFS: ®, = 80 MHz) provides a 
fine (s 2 1.875 m) but ambiguous range information over $,. 
Correct combination of the phase shifts ©, and ©, of both fre- 
quencies provides absolute and accurate range measurements 
within the specified range. 
R- CD 
NO 
  
  
  
   
Target 
Fig. 1: The two-frequency phase-shift method 
3. THE LASER RANGE SCANNER 
Hardware of the laser range scanner (Fig. 2) consists of two 
major components: the range measuring system and the beam 
deflection system. Both components operate independently 
from each other. They are connected via a control and moni- 
toring system. 
  
     
   
  
   
  
emitted ' L: avalanche 
laser light photodiode 
' * | receiver 
v received 
| laser light 
E rr > 
optics 
  
  
    
  
  
  
  
  
a a Wa LU 
ue = + id 
Fig. 2: Mechanical design of the range scanner 
Deviations from the continuously monitored eyesafe operation 
level or any deviations from the normal operation mode of the 
scanner, causes the camera to be shut off automatically. High- 
472 
speed da 
compute; 
The rang 
laser ran; 
cy unit ai 
sing. 
Laser H« 
The laser 
infrared ( 
tered froi 
micro-me 
The laser 
that emit 
Varying | 
modulate: 
60825) is 
of a grey: 
power rec 
as well as 
forms the 
with smal 
reflected 
optical pa 
f = 50 mr 
todiode as 
tral noise 
reflectanc 
High freq 
The high | 
laser diod 
quency-se 
light. 
For modul 
the frequei 
In order tc 
single 80 1 
ates modu 
current mc 
eliminatin; 
oscillator a 
The back-: 
tered in or 
channels, : 
ture mixer 
dynamic r2 
quadrature 
quadrature 
In-phase (F 
complex m 
tional to m 
guity inter 
dicates inte