Automatic Water Body Extraction from Point Clouds and Hydro-Flattening

ISAG2019 – International Symposium on Applied Geoinformatics, İstanbul, Türkiye, 7 - 09 Kasım 2019, ss.133-164, (Tam Metin Bildiri)

Nowadays, point clouds produced by LiDAR or photogrammetric methods are widely used in the generation

of high resolution Digital Terrain Models (DTM). However, DTMs produced from LiDAR and photogrammetric point

clouds cannot provide a smooth geometric representation in areas where water bodies exist. Therefore, in order to improve

the representation, hydro-flattened DTMs generated by adding breaklines to the water body boundaries in the point clouds.

The added breaklines creates a smooth surface over the water bodies. Determining and adding manually generated

breaklines is a time consuming process. Therefore, this process needs to be automated. In this study, hydro-flattened DTMs

have been produced by automatically determining the boundaries of water bodies using photogrammetric and LiDAR point

clouds. In order to extract the necessary information from point clouds with high accuracy, filtering process was applied

first. With this process, point cloud of the bare earth surface was obtained by removing the above ground objects in the

point cloud. In addition, outlier points such as dips and peaks, which are frequently encountered in the photogrammetric

point cloud, especially in the areas with water bodies, were determined by morphological operations and the height values

of these points were corrected depending on the points having smooth distribution around them. Then, the planarity and

curvature parameters were calculated from the point cloud to determine initial locations of water bodies. The boundaries

of the water bodies were determined by applying a segmentation procedure using these initial positions. In the final

processing step, hydro-flattened DTM was generated by using automatically determined boundaries of water bodies, and a

fixed height value was assigned to the points within these boundaries. Boundaries of lake, stream and sea were determined

by the proposed method which is applied in two different test areas. Accuracy assessment was performed by comparing

the determined boundaries with the aerial images and vector data of the water bodies of the study area. In the light of the

results obtained from the accuracy assessment, promising results were obtained with the proposed method.