Skip to Main Content
An innovative method to extract water feature from aerial-image is introduced in this paper. This approach extracts water feature from coarse to fine considering laser spectral bands of current existing airborne LIDAR systems and the spectral characteristic of these bands. Quad-edge based incremental inserting algorithm is used to construct the TIN (Triangulation Irregular Network) from LIDAR points. According to the triangulate features of different objects, area-analysis is performed to extract water triangles from TIN. Water triangles depict the water location of aerial-image. Then buffering is performed to extend the area of water triangle and to uptake the whole water-related points cloud data. Raster calculation is used here to obtain the rough water feature. Then, Mean-Shift algorithm is used to reclassify the rough water feature and to obtain the precise water. Finally, the feasibility of the approach is verified using comparison between two ordinary methods and the approach proposed in this paper.