3 T T 
e) : : : : : ; 
: i : contact : : : 
€ a Edere eed wires . o ay r5 
£ track : : : : : : track 
o : : : : : : 
2 service passage : : : : 
© ; ; : 1 
z05L IT T7 
o i E 
o : : 
B of 7 
=] | Li i | i 
300 350 
i L i 
0 50 10 150 200 250 
scanning angle B [deg] 
3c) Range errors resulting from filtering 
  
d) 
  
  
  
  
1800F oir Une cue een sa crée » = 
+ reconstructed range 
: - reference range 
1700 i i : 
0 5 10 15 20 25 
scanning angle ß [deg] 
3d) Reconstruction of track after filtering 
Fig. 3: Range errors depending on filter cut- 
off frequency (f, = 130 kHz) 
In order to evaluate phase €; and magnitude m; of P; and Q., a 
specialized Pythagoras processor is used. Nonlinearities and 
offset errors in the analog unit as well as physical effects of 
the APD cause deterministic errors with range and intensity 
measurements. To compensate these uncertainties, two hard- 
ware look-up-tables (d; = f(d.); Ad; — f(g.)) can be programmed 
for each channel independently after calibration. 
The resulting range value d (15bit resolution) is generated by 
means of a frequency-selective combination of phase shifts @, 
and @, of both channels. Absolute measurements within an 
ambiguity interval of d < 15m (LFS: 10 MHz) and high accu- 
racy of Ad = 0.457mm (HFS: 80MHz) are achieved. Resulting 
grey level g corresponds to the magnitude of the HFS with 12- 
bit resolution. 
Pixel oriented range data preprocessing, i.e. filtering software, 
coordinate transformation, and classification of single pixels 
as well as image processing algorithms are implemented on a 
transputer network based on seven T805 transputers. 
3.2 Deflection system 
For surface inspection of railway and highway tunnels, 360° 
profile measurement is required. Scanning of 360° profiles is 
achieved through the deflection of the emitted laser beam over 
a high-speed rotating mirror (Fig. 2). In order to fulfill the de- 
mand for gap-less tunnel inspection in combination with a 
speed of up to 5 m/s of the carrier vehicle (i.e distance bet- 
ween two consequtive profiles is less than 2.5 cm) the rotation 
speed of the mirror is controlled to 200 rps. 
Table 1 gives a survey of currently achieved system parame- 
ters. 
Resulting spiral profiles of the tunnel profile scanner are com- 
bined with respect to the actual sensor position. The final 
range image reflects geometric dimensions of the tunnel tube 
whereas the grey level image is used for visual inspection, sur- 
face classification, and documentation purposes. 
  
Laser head 
IR semiconductor laser diode (A = 810 nm) 
Hybrid APD photosensor 
Emitted laser power: 4.5 mW 
High frequency unit 
Two-frequency phase-shift method 
Modulation frequencies: 10/80 MHz 
  
  
Signal processing unit 
Reflectance (Pp/Pp): 2% ... 99% 
Depth of field: 0 -...15 m 
Accuracy of range data: up to 0.45 mm 
Sampling rate: 500 kHz 
Scanner system 
Beam deflection through rotating mirror 
Field of view: 360° profiles 
Spatial resolution: 2500 pixel/profile 
Profile imaging rate: 200 profile/s 
  
  
  
  
Tab.1: System parameters 
4. EXPERIMENTAL RESULTS 
In this section we describe a series of experiments designed to 
measure accuracy of the laser range scanner under different 
conditions and to compare it with the predicted theoretical 
values. Following Besl [4], accuracy, the difference between 
measured range and true range, and precision, the variation of 
measured range to a given target, are distinguished. 
We conduct experiments within structured test scenes by 
means of special test objects with standardized reflectances R 
(R € [2%, 5%, 20%, 50%, 75%, 99%]) in our laboratory. The 
area has both natural illumination that enters through win- 
dows, and artificial illumination generated by lamps mounted 
on the ceiling. These lamps will be referred to as spotlights be- 
cause they are powerful and somewhat directional. 
4.1 Accuracy 
To determine the accuracy of the range measurements is to 
identify the deviation between measured range and true range. 
474 
For a targe 
the scanne 
gin of the 
system. Ur 
r, but nonl 
in the anal 
lighting co 
5 
, range error (true range - measured range) [mm] 
  
Q 
2000 
Fig. 4: Sen. 
To determin 
of targets w 
the ambigu: 
at known d 
i.e. normal 
15m;0.5m 
ple of j z:] 
Experiment 
= max (ci 
range r are 
tables d, 0/8 
accuracy wi 
resulting rai 
flectance of 
Range drift 
To measure 
90°, R = SC 
range data i 
ting effects 
quency unit 
the first 10 1 
drift (Ad < 
other laser i 
shows good 
measuremen 
For characte; 
variation of 
bient lightin 
ce from the : 
target, and ai 
each of thes