Road Selection Based On Multi-attribute Decision-making Analytic Hierarchy Process

Map generalization is a hot research topic in spatial information processing.It aims to abstract spatial information to realize the generalization of maps from a large scale to a small scale.Due to the complexity of spatial information and contained ambiguity problems which cannot be described by mathematical models,map generalization has always been a challenge in modern cartography.The selection of geographical objects in maps is one of the operators of map generalization.The selection of lines is a more challenging subject than the selection of points and areas.Roads are major line features in maps.Road selection,a selective omission process of roads,is to retain the more significant roads that reflect the basic characteristics and distribution of roads,so as to achieve road generalization at different scales.The objective of this study was to develop a method for automated road network selection from a large-scale(1:50000)to a small-scale(1:200000)database.Among existing road selection methods,a method based on centrality performs best for this purpose.but it exposes problems that made it unable to meet cartographic requirements of all aspects,including:(1)The integration strategy of indicators is unable to reflect actual cartographic experience;(2)insufficient consideration is given to the contextual characteristics that affect the importance of roads;(3)it fails to maintain the density distribution of the generalized road networks;(4)the generalized road networks are not globally connected.To this end,this study proposes a road selection method based on Analytic Hierarchy Process(AHP).(1)Descript and construct a contextual characteristic indicator of roads assisted by POI data.The basic idea is to reclassify POIs and evaluate the importance of different categories of POIs with Delphi method,and the contextual characteristics are measured by the importance of POI categories and the quantity of POIs within a certain buffer.(2)Establish an AHP model with the subject of road importance.Firstly,betweenness centrality,degree centrality,length and the contextual characteristic indicator of roads are integrated in the criterion layer of AHP,whose weights are determined according to cartographic experience.Then,the indicator information of roads is analyzed and compared in the alternative layer of AHP.Finally,road importance is calculated by integrating the results of different layers.(3)Reflecting road density by the density of intersections and endpoints of roads,the density partition of a road network is performed based on Voronoi diagrams,and the thresholds of road importance are preset for individual partitions to interfere with selection processes.First,the Voronoi diagram of road intersections and endpoints is generated,thus to calculate the areas of Voronoi units.Then,~*was used to identify high and low-value aggregation areas of the Voronoi diagram at a certain degree of confidence,which are density partitions of a road network.Finally,importance thresholds are preset for the roads in different partitions to regulate the selection.(4)Define the weights of the edges in a line graph for connectivity maintenance.Firstly,the edge weights are determined by node importance calculated in the AHP model.Then a graph with largest edge weights is obtained based on the idea of minimum spanning tree under the premise of adding as few nodes as possible,which stands for the final result.The datasets of a road network and POIs in Hankou district,Wuhan city at a scale of1:50000 were used to verify the effectiveness of the proposed method.From visual and quantitative comparisons,it can be concluded that the generalized result at a scale of1:200000 by AHP-based methods better preserved the structure of the original road network compared with other methods.After the inclusion of the constraint of POIs,the selection result took into account the characteristics of surrounding facilities and the habitations.It also retained more roads with a high status in people's cognition,thus making the selection result more reasonable.In addition,the comparison between the results of manual selection and this method showed that the two results had a high degree of consistency.Although this method is able to improve the selection effect of the centrality method,it still requires manual intervention of experienced cartographic experts,and the generalized road network does not arrive ideal outcomes in all aspects.In the future research,the problems in road selection need to be refined,thus to design more comprehensive and automatic selection strategies.