International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
progresses, the height of more pixels is closer to zero (flat 
board). The overall accuracy of the algorithm was evaluated in 
each iteration by the percent of pixels that their height was 
within the required accuracy of 0.5mm. At iterations 1, 5 and 
11 the accuracy was 64%, 81% and 98% respectively. 
The same process was done to a real chicken filet. The number 
of iterations until conversion using the same parameters 
(€ ,@ ) was similar. The resulted projected pattern after the 
11? iteration was different than the one resulted in the white 
board experiments. As expected, because of the chromatic 
characteristics of the chicken filet, the span of the red band, is 
short across the pattern when comparing it to its span along the 
pattern in the white board experiment (Fig. 5). In the white 
board pattern it's below the linear line, which means slow 
change — good differentiation ability of the sensor, whereas in 
the chicken filet experiment most of it above the line (increase 
sharply in the beginning of the pattern) which means poor 
differentiation ability. 
  
  
. — linear 
| . Chicken filet > 
°¢" _ Whiteboard 
08r 
07! Jl 
06h 
05. 
04- 
| / 
037 ; 
02} f 
mi el. 
r Red range. 
% 20 40 60 80 100 120 140 160 
Figure 5. Color projection pattern after 11 iterations for the 
white board (blue line) and the chicken filet (green line) 
Figure 6 shows the chicken filet with the color pattern 
projected on it. The cyan line is the sampling line across this 
line the hue values are measured during the iterative process. 
Figure 7 shows the error rate of a complex topography body, 
similar to a chicken filet that was measured both by the 
proposed system and manually by a 3-D digitizer. The X and Y 
axis represent world coordinates in millimeters. The bar on the 
right hand side of the figure represents the error rate. 10,000 
sample points were measured across the body, 97.5% were 
within the error rate of -0.5mm. 
  
  
  
Figure 6. Chicken filet with pattern color projection on it 
during the iterative process 
300 
250 
200 
  
100 sun i 
50 
  
  
15 A 150 > oc 
Figure 7. Error rate of a measured body 
8. SUMMARY 
This rescarch apply photogrammetric / machine vision 
methodologies in the field of agricultural engineering. 
Customization of the methods was done as required by the 
special demands of the field. A method for generating color 
patterns of structured light was presented. This method can be 
used for industrial application when the texture and color of 
the measured object is uniform. Prototype of the suggested 
system was build and validated. The tests showed that 
measurement of chicken filet thickness is feasible in the 
accuracy level required for detecting bones fragments and 
other hazardous materials. 
   
Internat 
Caspi, I 
adaptive 
Intellige 
Geng, 7 
concept 
Enginee 
Karara, 
Falls C] 
Remote 
Mikhail 
Introduc 
New Yo 
Paakkar 
plates u: 
Salvi, J. 
review c 
evaluati 
Salvi, J. 
strategie 
Volume 
Sato, I. 
In: Proc 
Image C 
Californ 
Tajima, 
rainbow 
Recogni 
Wust, C. 
measure 
Applicat 
Yu, W.] 
1997. © 
moiré in