Automatic Detection of Gravel Bars in a River Channel from Airborne Lidar-derived Dtm

Airborne Laser Scanning or LiDAR data are widely used nowadays in river valleys for topography and hydro-morphology. However, the large data sets of unprocessed 3D point clouds require some challenging treatment and end-users may prefer to deal with DTMs (Digital Terrain Models) derived from LiDAR surveys. Without any complementary data (such as field survey or photographs) detecting position and shape of gravel bars in a river is a demanding task, especially if done manually. This paper presents a method for automatic segmentation of a river channel into distinct hydromorphological entities: water, gravel bars, banks, ... This method is based on image processing algorithms (region growing segmentation combined with morphological closing and altitude thresholding) in order to separate water from other elements present in the channel, based on the altitude information. This method is applied to a reach of the Arcen-Maurienne River, France, with alternate gravel bars. The only data source is the 0.25 m resolution DTM, derived from an airborne LiDAR survey. Results show that the developed method succeeds in automatically delimiting the main channel and in detecting gravel bars. It is then possible to get global information on the gravel bars such as location along the river or emerged surface area and topography.