5. CONCLUSIONS 
Applying the matching and estimation method of CV in MMS, 
broaden the research area of photogrammetry, it can be 
concluded from the experiment that have conducted: 
(1) Compare to traditional photogrammetry, the CV RO based 
on matrix have a more concise form, and is easier to 
programming, besides, this method suits for general heading 
angle. 
(2) Applying the excellent extraction and matching algorithm 
SIFT and RANSAC robust estimation method in vehicle-based 
sequential images processing, the experiments show this 
method has high stability. 
(3) Now the precision of RO based on CV is limited, but the 
precision is relative high. 
The quality of RO mainly relates to the quality of the images 
and matching algorithm. With the development of computer 
hardware and the popularize of parallel arithmetic, the time of 
extraction and matching will be shorter, this method is 
expected to be used in indoor visual navigation or provide 
initial RO value in position field, which has little require on 
time and precision. Moreover, after rigorous calibration, the 
quality can be greatly improved. 
RO is a fundamental question in photogrammetry and CV. RO 
based on vision is a method with low cost and relatively high 
reliability, so in traditional photogrammetry and CV, RO 
based on vision privilege over other method. But, the precision 
of the result by this method is not very high, and there are 
many factors that affect the result by vision method, including 
the quality of the image, the precision of camera calibration 
parameters, the quality of image matching, data processing 
and its optimization etc. Under the condition that the quality 
of the image is unchangeable, the quality of the result can be 
improved purely by improving the data processing method and 
algorithm, but the extent is limited. For example, the 
homologous points acquired by SIFT were not identified by 
correlation coefficient, the precision cannot be guaranteed. It 
is hard to guarantee large overlap (the overlap over 60%) (LI 
Jinwen and ZHAN Zongqian,2010), in order to meet the 
demand of multiple view matching and bundle adjustment. 
Besides, there are many uncertain factors in environment: 
these are important factors that influence the RO based on 
vision, with the improvement of relative technology, the 
quality of RO will be improved. 
Acknowledgements 
I would like to thank Beijing maggroup group for the vehicle 
based sequential image data of Landmark MMS. Also I 
appreciate the enthusiastic help of Cheng Xu, manager of the 
company, Doc.Wang Yanzheng and Liu Hua, postgraduate of 
School of Geodesy and Geomatics, Wuhan University. 
References 
[1] LI Jian et al., Best feature point matching method research 
and application in automatic relative orientation [J]. Science of 
Surveying and Mapping .2010. 4 
[2] Schenk, T., Toth, C., 1993. Towards a fully automated 
aerotrian-gulation system. Proc. ACSMrASPRS Annual Spring 
Convention, New Orleans, vol. 3, pp. 340—347. 
[3] D G. Lowe. Distinctive Image Features from Scale- 
Invariant Keypoints [J]. International Journal of Computer 
Vision, 2004. 60(2): 91-110 
      
   
    
    
      
    
     
     
     
   
   
   
   
    
    
    
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
   
   
  
  
  
  
  
   
  
  
  
  
  
   
  
    
[4]FISCHLER M ABOLLES R C.Random sample 
consensus:A Paradigm formodel fitting with applications to 
image analysis and automated cartography[J].Communications 
ACM,1981,24(6): 381-395. 
[5] McGlone, J.C., Mikhail, E.M., Bethel, J., Mullen, R. 
(Eds.), 2004.Manual of Photogrammetry, fifth ed. American 
Society of Photogrammetry and Remote Sensing. ISBN: |. 
57083-071-1. 
[6] Nistér, D., 2003. Preemptive RANSAC for live structure 
and motion estimation. Proc. of the IEEE Internationa 
Conference on Computer Vision, pp. 199—206. 
[7] ZHANG Zuxun. Digital Photogrammetry and Computer 
Vision [J] Editorial Board of Geomatics and Information 
Science of Wuhan University.2004. 12 
[8] Hartley, R; Zisserman, A. "Multiple View Geometry in 
Computer Vision." Cambridge University Press, 2003. 
[9] DING Meng; CAO Yun-feng. Method to obtain UAV's 
flight attitude based on computer vision [J]. Transducer and 
Microsystem Technologies.2007.26(11). 
[10] LI Jinwen,ZHAN Zongqian. Wide relief automated image 
stitching [J]. Chinese Museum.2010.01 
  
KEY 
ABS 
In th 
Euca 
decli 
distri 
the d 
Inves 
in. / 
analy 
varia 
crow 
1.1 
Ovet 
have 
bran 
want 
was 
incic 
Bro 
field 
The 
eval 
kno 
diffi 
insti 
Tim 
que: 
yeai 
Euc