Authors: G. Xiao, David Tien
Addresses: The School of Transportation, Wuhan University of Technology, Wuhan, Hubei, China. ' School of Computer Science and Mathematics, Charles Sturt University, Bathurst NSW 2759, Australia
Abstract: Detection of water areas on the land surface via aerial imagery is crucial for assisting land management. As near-infrared (NIR) energy tends to be absorbed by water, this property of low-spectral reflection is usually utilised in analysing water resources on land. However, the spectral reflection of shallow water varies significantly. It is difficult to distinguish such areas from the background by traditional land cover classifications. To solve this problem, this paper proposes an object-based classification approach for automatically detecting water areas from aerial imagery with red, green, blue and NIR bands. To overcome the problem of inadequate class definition in conventional region-based classifications, the water areas are divided into a number of classes, and a decision tree approach to select the features required for each class. Experiments show that the proposed approach has good capability to distinguish shallow water areas from other objects in wetlands.
Keywords: surface water detection; spectral feature; decision trees; aerial images; object-based classification; land management; shallow water; wetlands.
International Journal of Intelligent Systems Technologies and Applications, 2010 Vol.9 No.3/4, pp.218 - 227
Published online: 04 Nov 2010 *Full-text access for editors Access for subscribers Purchase this article Comment on this article