A Skeleton-based Density Partition Method for Evaluation of Road Network Selection
ISBN 978-85-88783-11-9
Authors
1Xiong, F.; 2Ren, C.; 3Wang, R.; 4Tian, J.
1SCHOOL OF RESOURCE AND ENVIRONMENT SCIENCE, WUHAN UNIVERSITY
2SCHOOL OF RESOURCE AND ENVIRONMENT SCIENCE, WUHAN UNIVERSITY Email: imr
3SCHOOL OF RESOURCE AND ENVIRONMENT SCIENCE, WUHAN UNIVERSITY
4SCHOOL OF RESOURCE AND ENVIRONMENT SCIENCE, WUHAN UNIVERSITY Email: yutaka-2010@163.com
Abstract
Road density is a widely used index in the analysis of ecological effects, road network planning, delimitation of urban areas and road network selection. To evaluate results of road network selection, many indexes are being used, such as using similarity to evaluate the length similarity of road networks before and after selection, using connectivity to measure the connectivity of retained road networks, and using visual inspection to visually check whether the road network pattern is preserved. However, little attention has been paid to use density to evaluate results of selection. Although mesh-density based approach and the recently proposed combination of stroke-mesh approach can retain the road network density difference, these methods did not use density to evaluate results of selection. This paper proposes a road density partition method based on skeleton partition of road network and the idea of spatial autocorrelation. Firstly, constrained Delaunay triangulation is applied to the road segments and skeletons are generated. Then, approximated Voronoi polygons of each road segment are formed based on these skeletons. Secondly, Getis-ord Gi* is used to identify statistically significant spatial clusters of high and low values of Voronoi polygon’s area. Finally, the neighboring Voronoi polygons are aggregated based on the statistically significant spatial clusters of high and low values. The road network of Hong Kong at the levels of 14, 13 and 12 in the Google Maps were used as experimental data for the evaluation of road network selection. The result shows that the densities in the experimental partitions by and large match the road network density in reality, and well reflect the road density contrast before and after selection. Additionally, the result shows that this method is better compared with grid density method.