COORD. y , 
T 23 4 5 
—— AZUL MEDIO 
  
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 
AZUL CLARO ———INFRA ROJO IR FILTER —- SIN FILTRO ——— VERDE OSCURO 
Figure 7. Principal point y, (in pixels). 
  
1 2 3.4 
—— AZUL. EDO - 
minmimmmmmmm mmmmmmm 
hbbiose iHd mis rca re 
AA AAA NO ROO 
mmmmm 
1.2 3 4 
—— AZUL, MEDIO 
COOR. X, 
1-23 4—5 
5.6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 
AZUL_CLARQ === INFRA ROJO # IR_FILTER sme SIN_FILTRO swe VERDE_OSCURO 
Figure 8. Additional parameter XK. 
  
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 
AZUL. CLARO ——- INFRA ROJO ese IR_FILTER ——— SIN_FILTRO =e VERDE_OSCURO 
Figure 9. Additional parameter K>. 
  
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 
—— AZUL MEDIO AZUL CLARO «—— INFRA ROJO IR. FILTER —— SIN FILTRO ———— VERDE OSCURO 
Figure 10. Projection centre X; (m). 
  
BBrrsrsnref OR bb osoosooo 
ANOBMUONOOOSNUBOONDOO+ 
12 3 4.5 
—— AZUL. MEDIO 
   
  
i 
3 
c] 
| 
| 
i 
  
€ 
E 
T 
| 
E 
Il 
| 
| 
T 
i 
i 
  
: CURE: mE] : 3 
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 
6 7 8 9 10 
AZUL. CLARO ——— INFRA ROJO IR. FILTER —— SIN. FILTRO -—— VERDE OSCURO 
Figure 11. Projection centre Y; (m). 
    
hI) NONI OD PILI UOS AS! 
  
123456 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 
—— AZUL MEDIO AZUL CLARO ——- INFRA ROJO —IR FILTER—— SIN FILTRO ——- VERDE OSCURC 
Figure 12. Projection centre Z; (m). 
Fight object points were measured for every image. The 
methodology and the environmental and operating conditions 
were was always the same. The mean-square error of unit 
weight a posteriori of the least-squares adjustments was 1.5-2.7 
pixels. 
5. DISCUSSION 
Data obtained from this study show that the stability of video 
cameras is very poor. Whatever parameter is analysed, the 
variability of the data is higher than expected. 
The nominal principal distance 8 mm, according to the pixel 
size should be 1081 pixels, but after the study, the median is 
1139.6, the mean 1144.3, and the standard deviation 28.2 
pixels. Furthermore, the principal distance is surprisingly 
independent of the wavelength (Fig. 5). In fact, in theory, it 
should be bigger when the wavelength increases. 
Regarding the principal point (Figs. 6 and 7), the relationship 
between image space and object space is strong in both 
directions. The cross-correlation between x, and X; is —0.72, 
and between x, and Z;, 0.70 (because of the terrestrial system); 
between y, and Y, is —0.74. The mean x, co-ordinate is 
approximately 316.4, i.e. —3.6 pixels from the image centre, and 
yo 230.8, 1.e. —9.2 pixels. The standard deviations of both values 
are, 0,,=22.1 and o,,-18.4 pixels, respectively. 
The strongest correlation between parameters is between the 
principal distance f and Z;, offering a value of 0.95. However, 
the relationship among f and the radial distortion parameters 
(Kı,K,) is very low, 0.27 and —0.48, respectively. Obviously, 
the latter parameters are highly correlated, -0.95. The 
correlation between additional parameter K, and the principal 
point is much stronger than with the parameter K, (1.5-2.1 
times). 
The two internal parameters Æ, scale factor in y-co-ordinates, 
and ó, shear (orthogonality error) between the x- and y- axes, 
show different behaviour. The former offers a median value of 
0.0027 and a standard deviation of approximately the same 
value, 0.0021. The latter gives consistent values, the median 
equals 1.02 and the standard deviation is 0.0018, so, the affinity 
correction should not be omitted in further studies. 
Besides, the external orientation parameters also show 
additional fluctuations on their values, though there are not 
isotropic. Standard deviations in projection centres range from 
0.11 m in Y direction (Fig. 11) and 0.28 m in Z direction (Fig. 
12) to 0.77 m in X direction (Fig. 10). Similarly, the standard 
deviations in orientation angles fluctuate from 0.5? in y to 1.15* 
in o and 1.21* in c. 
The problem when using this type of video camera is to know 
when your data are suitable or not for your processing. 
Obviously, series of the same scene guarantee right values 
-402— 
takin, 
and a 
The 
came 
captu 
Resul 
video 
the v. 
are m 
non-r 
geom 
prom