The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008 
543 
occlusions by vehicles, but the shadow and occlusion is not 
serious. In the procedure of tracking, there are 8 times the 
tracker deviating the path, and then the thread is ceased as soon 
as possible by the operator. There are 6 times prompts notifying 
the user that the tracker is no longer suitable and it needs 
manual plotting. And there are only 76 manual input points, and 
the whole process takes 543 seconds. But if the operator want 
manually plot all path point with the same precision as the 
tracker, it needs 1108 inputs and it takes 776 seconds. In 
general, the quality of the result of manual and semi-automatic 
plotting is equivalent, since the operator supervises the results 
of the semi-automatic system and failures are edited online. On 
average, the geometric accuracy is comparable, too. We also 
test our tracker on many other grey scale images with different 
resolution varied from 0.2 to 2.5 m, and the results are similar 
in manual input saving about 90% and time saving about 30%. 
The result shows that our tracker is quite robust when the 
photometric property of same road segment changes suddenly, 
and when the tracker reaches to the junctions and it will go on 
without stop. And the tracker can detect the road centreline of 
the roads in any orientation with moderate curvature accurately, 
and also works successfully for roads have some obstacles 
caused by shadow and occlusion. 
Fig. 4 Semi-automatic extracted road network on 
Quickbird image (a) Overview (b) Local result. 
6. CONCLUSIONS 
The proposed tracker based on ATS is very robust, because it 
makes best use of the road characteristics on high-resolution 
imagery. Our algorithm employ parabola equation to fit the 
trajectory of the road and to predict the road position and 
moving direction and to judge whether the new added road 
point is right by check the curvature change, it also utilize 
compactness coefficient to evaluate the aptness of itself to go 
on tracking, so the algorithm has some ability of higher-level 
reasoning. The current limitations are that the algorithm may 
not work on the road cast by much shadow and occlusion in 
complex scenes, it can’t judge the validity of input seeds, it can 
only track long ribbon roads on grey scale imagery, and it need 
more computing times. These limitations are currently being 
examined now. The main contribution of this paper is that it 
employs angular texture signature semi-automatically extract 
road with precise results. 
REFERENCES 
Bajcsy, R., Tavakoli, M., 1976. Computer recognition of roads 
from satellite pictures. IEEE Transactions on Systerms, Man, 
and Cybernetics, 6(9), pp. 623-627. 
Baltsavias, E. P., 2004. Object extraction and revision by image 
analysis using existing geodata and knowledge: current status 
and steps towards operational systems .ISPRS Journal of 
Photogrammetry & Remote Sensing, 58, pp. 131-151. 
Barzohar, M., Cooper, D. B., 1996. Automatic finding main 
roads in aerial images by using geometric-stochastic models 
and estimation. IEEE Transactions on Pattern Analysis and 
Machine Intelligence, 18(7), pp. 707-720. 
Baumgartner, A., Hinz, S., Wiedemann, C., 2002. Efficient 
methods and interfaces for road tracking. International Archives 
of Photogrammetry and Remote Sensing, 34(3 B), pp. 28-31. 
Benz, U. C., Hofmann, P., Willhauck, G., et al., 2004. Multi 
resolution, object-oriented fuzzy analysis of remote sensing 
data for GIS-ready information. ISPRS Journal of 
Photogrammetry & Remote Sensing, 58, pp. 239- 258. 
Doucette, P., Agouris, P., Stefanidis, A., 2004. Automated road 
extraction from high resolution multi-spectral imagery. 
Photogrammetric Engineering &Remote Sensing, 70(12), pp. 
1405-1416. 
Gardner, M. E., Robert, D. A., Funk, C., Noronha, V., 2001. 
Road extraction from AVIRIS using spectral mixture and Q-tree 
filter techniques. In: Proc. AVIRIS Airborne Geosciences 
Workshop, Pasadena, California, 
URL:ftp://popo.jpl.nasa.gov/pub/docs/workshops/01 docs/toc.h 
tml (last accessed : April 20,2007). 
Geman, D., Jedynak, B., 1996. An active testing model for 
tracking roads in satellite images. IEEE Transaction on Pattern 
Analysis and Machine Intelligence, 18(1), pp. l-14.Gruen, A., 
1985. Adaptive least square correlation-A Powerful Image 
Matching Technique. South African Journal of Photogrammetry, 
Remote Sensing and Cartography, 14(3), pp. 175-187. 
Gruen, A., Li, H., 1995. Road extraction from aerial and 
satellite images by dynamic programming. ISPRS Journal of 
Photogrammetry and Remote Sensing, 50(4), pp. 11-20. 
Gruen, A., Li, H., 1997. Semi-automatic linear feature 
extraction by dynamic programming and LSB-Snakes. 
Photogrammetic Engineering and Remote Sensing, 1997, 63(8), 
pp. 985-995.