rj Field A2 A2 A3 A3 
[regain — à [m | m [ m | 5 
| Perioés | Field 
8 | s 
puce] BY : Y Y Y Time t 
| Video Signal | 
A = Field Field Field Field Field Field 
A B A B A B 
[, 20ms ,| 
Frame 1 Frame 2 Frame 3 
| 40ms | 
—— | 
Heckel 
Fig. 6: Temporal interdependence of video si- 
gnal and integration periods (IL-sen- 
sors) 
Fig. 6 shows the temporal interdependence between 
the actual video signal and the integration period 
to which it belongs. The integration period for the 
video signal of a field are the two preceding 
fields. For example (see Fig. 6), the information 
of field B1 was integrated during the read-out pe- 
riods of fields Al and A2. At every time two conse- 
cutive fields are exposed simultaneously. Perfect 
exposure of the profile (with full spatial resolu- 
tion) must take place during periods of simultane- 
ous exposure of two fields (for example fields Bl 
and B2) belonging to the same frame (for example 
frame B). For every frame the period of simultane- 
ous integration of both fields is the read-out pe- 
riod of the second field of the preceding frame. 
Exposures during this periods will be integrated in 
both fields of the next frame. One scan in this pe- 
riod will be detected with full spatial resolution. 
Hence the following aspects are essential for syn- 
chronization of IL cameras to systems with scanning 
illumination or flashlight illumination: 
- The two fields of a frame are read out in an 
interlaced modus, but each sensor element in- 
tegrates charge packets for a full frame pe- 
riod. 
- During the read-out period of the second 
field of a frame both fields of the next fra- 
me are integrated simultaneously. If the 
scanner (resp. flashlight) is active during 
this period of approximately 20 ms, the next 
frame will contain one exposure with full 
spatial resolution. 
= IL cameras will show only very little smear 
effects, even for accidential exposure during 
transfer periods. 
For this reasons interline-transfer CCD cameras 
will be preferable detectors for use in systems 
with scanning or flashlight illumination. 
Synchronization in the presented measuring system 
The selected CCD camera (SONY XC 77 CE) is based on 
an interline-transfer sensor. Exposure of the ob- 
ject must take place during the integration period 
of any second field. The next frame of the video 
signal will contain the desired image and has to be 
captured by the frame grabber of the host computer. 
A reproducible relation between camera cycles, 
scanner motion and frame grabber is assured with 
the following procedure: 
23) Generation of the scanner program: 
Starting with user-defined coordinates for starting 
and final point of the scan vector the host compu- 
ter calculates the scanner program in such manner, 
that the laser beam will move along the scan vector 
during the integration period of 20 ms. Different 
vector lengths will cause varying exposure of the 
profile. Later on this drawback will be abolished 
by controlling the laser power from host computer. 
2) Programming the scanner unit: 
The host computer programs the scanner electronics. 
The mirrors move to the starting point, the scanner 
unit stores the coordinates of the final point and 
waits for the execution command. 
412 
3) Svnchronizing with camera timing: 
The host computer waits for the next vertical blan- 
king pulse (i.e. the beginning of the next full 
frame). Then it gives execution order for the scan- 
ner unit. 
4) Scanning: 
The scanner waits for the beginning of the second 
field (experimentally determined delay time). Then, 
after acceleration of the mirrors, the laser is 
switched on and the laser beam moves along the scan 
vector. The scanner electronics independently cal- 
culates flat-field corrected 'micro-steps' for the 
mirror motion. 
5) Image acquisition: 
The host computer grabs the next frame of the came- 
ra signal for evaluation in the frame storage. 
With this procedure a correct exposure of a full 
frame is assured independent from length or posi- 
tion of the scan vector. Remember that the laser 
beam is switched on only for an exposure time of 
20 ms, which reduces possible dangers of the invi- 
sible laser diode beam significantly. 
EVALUATION OF PROFILE POSITIONS IN COMPUTER IMAGE 
The position of the profile is evaluated in each 
vertical column (or horizontal line, depending on 
the direction of the profile) of the computer image 
with subpixel accuracy. Thus each column delivers 
one measuring point and each leg of the profile 
composes of about 200-250 data points for profile 
positions. Linear regression for each leg yields to 
predictions for the straight lines with statistical 
improved accuracy. For absolute measurings of ben- 
ding angles the conclusion from profile's positions 
in computer image to 3-D coordinates of workpiece 
surface must be drawn. 
Gaussian interpolation algorithm 
Subpixel accuracy profile position in each column 
is achieved using interpolation algorithms. First 
the computer searches in each line resp. column of 
the computer image for the pixel with maximum in- 
tensity, assuming it to be an approximate position 
of the profile. An interpolation function is fitted 
through the intensity distribution of this pixel 
and its direct neighbours. The calculated center 
position of the interpolation function is interpre- 
ted as position of the profile in this line resp. 
column of the computer image. Hence the profile is 
now represented by 512 data points with subpixel 
accuracy. 
A lot of interpolation algorithms have been discus- 
sed by different authors (see for example [6], 
[7], [13], [14], [15]). Interpolation algorithms 
based on Gaussian distribution curves are quite ob- 
vious, because the peak of almost every laser beam 
intensity distribution shows approximately Gaussian 
character. Eg. (1) denotes this assumption by 
2 
IG) - A:exp(- C27) (1) 
where: 
I(x) Gaussian intensity distribution 
A,B,C amplitude, width and center position 
of the Gaussian distribution curve. 
The information from at least three pixels is ne- 
cessary to calculate the three parameters A,B,C of 
the Gaussian interpolation curve. Let these three 
pixels be the pixel x, with maximum intensity I, 
and two of its neighbours (x,,I,) and (xy,I,) with 
1,<I, and I,<I,. Essential for calculating profile 
positions is only the center position C of the 
Gaussian distribution curve, which is determined by 
Eq. (2) 
with 1: Lint, /1,) 
_ (Xp-Xo) L-R? inn) (L*R) (2) 
2 L-R 3 
and Roma Ko 
Xp” Xo