Abstract: Aimed at the resolving problems in selecting shallow sounding features by manual operations, such as low efficiency and easy omission, a new method for selecting features based on the slope-relationship from soundings in digital charts is put forward in the paper. By constructing a weighted Delaunay triangular network for soundings, the slope-relationship between different soundings can be confirmed by the spatial rule, and the latent lines from the soundings can be distilled, thus enabling automatic shallow sounding feature selection. Experiemental results show that this new method is an effective way to improve efficiency when distilling feature shallow soundings and avoid the omissions.