A Shape Similarity Assessment Method for Linear Feature Generalization

Focusing on the shape similarity aspect in linear feature generalization's quality assessment problem, this paper presents a novel shape similarity assessment method for linear feature generalization. By introducing the constrained Delaunay triangulation and convex hull, the two-side root-bend series are generated. After that, the bend trees are constructed on every root-bend to form the two-side bend forest shape representation model, in which every bend on every level is represented by a triangle. Based on this model, together with the geo-location of the linear feature, the shape similarity between the linear feature before and after generalization is assessed. Experimental results show that, this method can distinguish shape features of different levels effectively, thus achieving the same result as subject shape cognition, and can identify quality problems caused by linear feature generalization.