Feature based Pointwise method 
a b e d 
Xo 6.36 18.69 A-12.33 6.43 10.91 A- 4.48 
Yo 6.36 18.69 A=12.33 6.43 10.91 A= 4.48 
Zo 1.58 158 A= 0.00 1.94 2.74 A= 0.80 
®. 0.58 174 A- 116 0.538 1.16 A= 0,58 
© 058 174 A= 116 0.58 1.16 A= 0.58 
« 095 074 À 049 0.25 035 A= 0.10 
  
Table 2 Weight coefficients of exterior orienta- 
tion parameters. Datum used in the 
determination; four feature lines and 
ground points. /Mets91/ 
From Table (2) it can be seen that, when the ob- 
servations are made in a very narrow interval, 
same accuracy cannot be achieved with feature 
based methods as with pointwise methods. With 
a rational selection of observations there are no 
difficulties to determine the parameters as accu- 
rately as with pointwise methods. We have to 
remember that by using feature based methods, 
we are able to make more observations along the 
line much more easily than by measuring more 
ground points in pointwise methods. Obviously, 
the increased number of observations improve 
the precision of estimation. 
Same kind of experiment was accomplished with 
splines. Because the photogrammetric treatment 
was based on the collinearity condition, once 
again there is a different number of observa- 
tions. In the feature based method 32 and in 
pointwise method 10 points were observed. The 
distribution of points is illustrated in Figure (9). 
  
  
  
a) 21756 b) 
  
  
  
  
  
  
  
  
  
  
The distribution of observations in 
a pointwise approach a) and in a 
feature based method along a 
spline b). /Mets91/ 
Figure 9 
  
Feature based Pointwise method 
Xo 7.84 9.13 
Yo 9.32 9.16 
Zo 1.52 2.13 
® 1.29 1.35 
© 1.15 1.34 
K 0.38 0.34 
  
Table 3 Weight coefficients of exterior orienta- 
tion parameters. Datum used in the 
determination; spline feature and 
ground points. /Mets91/ 
506 
In Table (3) the precision of estimation is given 
in a same manner as previously. 
4. CONCLUSIONS 
In the previous presentation we have shown that 
feature based methods are competetive with the 
pointwise approach. It can be stated that hardly 
any object line or curve could measured with the 
same precision as the corresponding accuracy of 
ground points. But the main thing in feature 
based methods is you can make much more ob- 
servations along a linear feature and thus mak- 
ing the determination of exterior orientation at 
least as accurate as with the pointwise proce- 
dure. In the case of intersection, the accuracy of 
pointwise methods was easily achieved. 
Assuming that with real, practical projects we 
can get same results, feature based methods will 
be a real, considerable alternative for the stand- 
ardisized procedure of map revision. Even with 
the semiautomatic approach, savings of time and 
money with be noticeable and speak on behalf of 
the use of this new method. We are convinced 
that in few years numerical databases, with 
numbers of accuracy, will be quite common pro- 
viding thus good potentials to imply feature 
based methods in practical map production. 
5. REFERENCE LIST 
/Heik92/ Heikkinen, A.J., 1992. Linear featu- 
res in photogrammetry (in Finnish). 
Master's Thesis, Helsinki University 
of Technology, Espoo, Finland. 
/Mets91/ Metsämäki, S., 1991. Basics of Featu- 
re Based Photogrammetry (in Fin- 
nish). Master's Thesis, Helsinki Uni- 
versity of Technology, Espoo, Finland. 
/Mula89/ Mulawa, D.C., 1989. Estimation and 
Photogrammetric Treatment of Lin- 
ear Features. Dissertation, Purdue 
University, Michigan, USA. 
/MuMi88/ Mulawa, D.C., Mikhail, E.M., 1988. 
Photogrammetric Treatment of Lin- 
ear Features. In: Int. Arch. Photo- 
gramm. Remote Sensing, Comm. III, 
Kyoto, Japan.