Airborne laser scanners (ALS) usually rely on a near-infrared light beam which is absorbed by water. This produces empty areas with no points in the LiDAR dataset (gaps, laser shot dropouts). Detecting the boundaries of bodies of water in a LiDAR dataset can thus be seen as the identification of boundaries of empty areas. The method for the identification of water bodies could be used to update and supply the Database of Topographic Objects (BDOT10k). The x, y co-ordinates of ground laser points of the previously classified LiDAR point cloud of the 4 points/m² nominal density were used to detect bodies of water. The automatic identification of bodies of water was performed by the means of an α-shaped algorithm. Numerical experiments were conducted for 16 tested sites, which were bodies of water. The accuracy of boundary identification was evaluated by comparing the results with those seen on orthophotos with a pixel size of 0.10 m. Based on the maximum deviation values of the measured results, it has been shown that bodies of water were identified on average with 95% accuracy and the boundaries of 62% of the tested sites were delineated with 100% efficiency. Furthermore, the studied algorithm has a featured mechanism that enables it to eliminate single, random points distributed on the surface of a body of water. The proposed method can be used as an additional source of BDOT10k, especially for bodies of water whose banks are covered with vegetation which are difficult to identify on orthophotos.

13-26