Figure 7 The Superposition result 
pwwfi w$m wm 
From the above experiment we know that :1) if we can fully 
understand and describe geometry characteristic and the 
spectrum characteristic of the objects information, and select 
appropriate structure element, , we can extract majority road 
information that satisfies certain size and geometry by the 
mathematics morphology method., especially as the road 
geometry shape is complex changeable ,detail numerous and 
diverse;2) the improvement method of region growth - seed 
growth, which eliminates burr in the road median line and the 
non- road shoreline and more effective improves extraction 
precision;3)in actual production process, the people need extract 
other objects in the remote sensing image , such as the rivers 
network, the building, the inhabitant and so on, these objects all 
have their own unique geometry and the spectrum characteristic. 
Therefore, the road information extraction thought may expand 
to other remote sensing image objects extraction. 
At the same time , we also find that, because of the limited of 
the mathematics morphology theory and the complex of road 
information in the remote sensing, there also has its deficiency 
in the experiment: 1) As the extraction objects and the noise has 
the same geometry and spectrum characteristic, the 
morphology method is no longer effective; 2) Different objects 
processing needs different structure elements, However, 
currently, structure element selection doesn’t unification theory 
basis, in this paper the structural element selection has been in 
the certain experience foundation, through experimental 
analysis definite.;3) the extraction result through mathematics 
morphology processing is decided by the image segmentation 
result in the great degree therefore, if we use a more highly 
effective segmentation method, we may unify it with the 
existing some other methods, especially the context information, 
or the texture information method ,we would increase the 
precision. 
This paper gives priority to mathematics morphology and gives 
assistance to the seed growth method, compared with the 
independent mathematics morphology or seed growth, the 
extraction precision has been improved, this has a vital 
significance for the research area such as the targets extraction, 
the GIS data regarding, automobile navigation, urban planning, 
and map digital and so on. Using many kinds of extraction 
methods to unify, the multi-scale multi-temporal multi-spatial 
resolution image fusion, extracting the high precise road 
information will become the hot research spot in the future. 
REFERENCES 
[1] Jin X., Davis, C.H. 2004. An integrated system for 
automatic road mapping from high-resolution mufti-spectral 
satellite imagery by information fusion, Elsevier Science 
Information Fusion, pp. 2 
[2] Shi Wen-zhong,Zhu Chang-qing,WangYu. 2001.Road 
Feature Extraction from Remotely Sensed Image: Review and 
Prospects[J].Acta Geodaetica et Cartographica Sinica, 30(3), 
pp.257-262.