International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 li 
  
descriptive statistics give an indication of the spread of the data Histograms from plane fit residuals — plane rotated with T 
and therefore allow any discrepancies in the data between respect to the scanner b 
experimental conditions to be detected. For the purpose of this ; ; : 
paper only the white, Neutral 8 grey and black from the Experimental results from the case where the colour chart was 6 
greyscale patches and blue, green and red from the primary rotated with respect fo the axis of the” scanner demonsirate m 
colours will be discussed. significant discrepancies. Figure 7 describes variations in the e 
histograms for the white and black patches. Data are also th 
Figure six shows a typical set of histograms for the white, summarised in Table 2. at 
Neutral 8 grey and black residual values following plane fitting : : re 
at position one with the planes orientated normal to the scanner. Whigham ni 
Table 1 documents a range of sample statistics connected with ui iq fü 
these data and also those for the Red, Green and Blue colour M f me is as 
patches. M1 | | Fast So ses in 
12 as 
Histograms for all of the patches scanned with the chart Zo} 
orientated normal to the scanner exhibit very little variation. 2 e) C 
There is a slightly increased standard deviation for positions * ol T 
seven and eight, which may be due to ambient light levels as the 4 p: 
chart was scanned outside for these positions. Of greatest i i 
importance is that the standard deviation of the residuals normal ; | Su 
  
to the planar surface was of the order of 2 to 3mm, with minima et 4 20 21 454-8100 E 
and maxima of up to 12mm, for each individual surface patch. | DET 
  
  
   
   
       
     
      
    
    
      
      
    
     
    
   
      
   
    
    
     
    
      
   
          
   
       
    
       
     
     
     
       
     
  
     
    
   
   
   
   
  
   
   
  
  
     
     
      
        
     
  
     
       
      
        
     
        
     
   
     
    
    
      
     
    
     
  
   
    
   
  
  
   
  
   
   
        
     
    
       
      
     
  
     
    
     
   
        
        
     
       
  
        
  
      
        
18 ——— White a 5 
Black Patch ea 
16 + | - — — - Neutral 8 Te Fro pa 
44 em Black 16 | Pos 1,20 deg gr 
i EE ac 
12 ie at 
d 3 401 bu 
d $ 4] ap 
A e of 
co 
s rel 
2 
va 
2 IM 
HO RB wh 2 0 2° 4 ‘6-8 10 no 
Distribution from Mean po 
Distribution from Mean (b) 
Histograms of variation in data distribution from the 
Fig.6 White, Neutral 8 and Black patch distributions mean at 0% 20°, 40° and 60° to the scanner 
at position 1 (4.08m) Fig 7(a) White Patch and 7 (b) Black Patch 
An f-test was used to determine whether two populations taken Table 2: Descriptive statistics (mm) for the white, Neutral 8 
from the same patches at different ranges were significantly grey and black and colour patch data at position one. 
different. The samples were tested using a two-tailed test at the Chart ali orthogonal and 40 de to the scanner. 
95% confidence level, the null hypothesis being that the Position Orthogonal 40 rees 
variance of the two samples are equal. Results of the f-test Patch N 8 W |NS8 
demonstrated no significant difference between the same colour Std Dev 2.6 1.7 F 
patches at the different ranges when positioned orthogonal to Min -82 53 [48 
the scanner. For example, there was no significant difference Max 88 6.8 1.6.8 
between the points defining the white patch at position one Patch B G B 
(4.08m) and position five (7.75m). This shows that the response Std Dev 2.6 18.119 
of the scanner to different orthogonally positioned colour Min 21-82 54.163 
patches does not change over the short range tested. Max 72 | &8 59 1.66 
Table 1: Descriptive statistics (mm) for the white, Neutral 8 
grey and black patch data at positions one, three, five and six As expected from the scanning geometry the number of points 
Position 7 (4080mm 5 7750mm captured (area under the curve) decreases with an inereise in 
incident angle. The charts show that the normal distribution 
Patch - NS V NS improves as the incidence angle is increased. The effect is most 
Std Dev | 2.5 | 3-6 32 |. 3.6 pronounced with patches that have greater reflectivity at the Fig 
Min -71 + =9.0 -7.0 | -9.0 wavelength of the laser. 
Max 9.0 97190 
Patch G B G B F-test computations indicate that the observed changes in The 
Std Dev | 24 | 3.2 30 | 32 histogram shape with rotation angle are significant. For bet 
Min -82 | -10 -10 example, the white patch at 0° is significantly different to white qua 
Max 72 | 92 9 at 40° at the 90 per cent significance level. This verifies the pat 
observed differences between the histograms where it appeared dec 
that the shape of the data distribution improved when each patch e 
e 
1034