The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
249
Figure 9. LiDAR image’s PFF recognition result
•Artificial river • Highway • Natural river •Other terrain
Then the areas of three terrain categories are segmented from
RS image, which is showed in figure 10.
Figure 10. RS image’s manual interpretation result
•Natural river •Highway •Artificial river
The comparison shows that the result acquired with the profile
factor algorithm is close to from high-resolution RS images by
manual interpretation.
Figure 11. RS and LiDAR recognition results’ comparison
•Natural river •Highway •Artificial river • Other terrain
The rivers can be distinguished with highways, and the artificial
rivers also can be distinguished from natural ones. Artificial
rivers’ area is taken as criterion, and the extracted rivers’ area
occupies about 84% of the related area interpreted from RS
images.
3.3 Analysis
The proposed algorithm based on profile factor adds criterion
up to the automatic identification of rivers, but this can not be
applied in all practical applications. The rivers’ straightness can
be used as the additive discrimination condition for artificial
river and natural river, highway and natural river.
At the same time high-resolution remote sensing images and
LiDAR data can also be fused to increase the accuracy of
automatic recognition. During related software development,
various proper factors should be used and integrated to enhance
the accuracy and reliability of automatic identification.
4. CONCLUSION
This paper explores a new river automatic recognition method
combined with terrain morphological distribution, based on the
specialties of LiDAR ranging data. And the work supplements
recognition function to feature extraction modules and supplies
appropriate scheme for automatic application.
ACKNOWLEDGEMENT
This work was supported by the Wiser Foundation of IDC -
Peking University (No. W08SI01). The authors also wish to
thank LSU CADGIS Research Laboratory and Louisiana Oil
Spill Coordinators Office of America for distributing laser
scanning data and high-resolution RS images of Abbeville
district.
REFERENCES
Abe, K., 1994. Thinning of gray-scale images with combined
sequential and parallel conditions for pixel removal. IEEE
Trans, on Systems, Man, and Cybernetics, 24(2), pp. 294-299.
David, J., 2006. Extraction of tidal channel networks from
airborne scanning laser altimetry. ISPRS J. of Photogrammetry
6 Remote Sensing, 61, pp. 67-83.
Guido, M., 2004. Robust circle detection using a weighted MSE
estimator. Proceedings ofICIP’04, pp. 2111-2114.
Hong, W., 2006. Unsupervised segmentation using Gabor
wavelets and statistical features in LiDAR data analysis.
Proceedings of ICPR '06, pp. 667-670.
Jonathan, H., 2007. The generation of river channel skeleton
from binary images using raster thinning algorithms. GISRUK
07 Proceedings, pp. 101-107.
Lam, L., 1992. Thinning methodologies - a comprehensive
survey. IEEE Trans. Pattern Analysis and Machine Intelligence,
14, pp. 869-885.
