Research On Urban Road Selection Methods Based On Knowledge Graph

As a basic component of the geospatial database,urban roads are important infrastructure,the basis for expression,analysis,planning and service of urban spatial activities,and constitute the basic skeleton of urban form.The road network carries human production and life,and to a certain extent the accessibility of the road network affects the spatial distribution of the city.With the rapid development of China’s economy and 3S technology(RS,GIS,GPS),the collection,modelling,updating and service of road data has developed rapidly,and in recent years,the development of new industries such as unmanned vehicles,intelligent transportation and smart cities,the demand for different scales of road data is skyrocketing,so take road synthesis to obtain different scales of road data is a necessary means,and the selection as road The selection as one of the operators of the road mapping synthesis is usually the first technical step of the road synthesis.Roads have a variety of shapes,grades and types,as well as the complexity of spatial relationships and other characteristics,resulting in the automatic selection of roads is the most important and difficult task of cartography at this stage,the previous way of manual selection takes a long time,seriously affecting the current status of geospatial data,so there are many scholars on the automatic selection of roads to carry out in-depth research,explore new theories,new methods,aimed at improving the road Therefore,many scholars have conducted in-depth research on automatic road selection,exploring new theories and methods,aiming to improve the degree of automation of road selection and speed up the frequency of map updates to meet the growing needs of people,but at present,the degree of automation of road selection is still low,and the selection index weights and thresholds are too much influenced by human subjective factors.In order to improve the degree of automation of road selection and its selection accuracy,this paper proposes a knowledge graph-based urban road selection model based on graph theory,knowledge graph theory,and graph convolutional neural network method,which makes comprehensive use of semantic features,geometric features,spatial features of roads,as well as the selection experience of cartographers,and strives to reduce the participation of human factors to achieve end-to-end automatic road.The method aims to reduce human involvement and achieve automatic end-to-end road selection.The main research work and results of this paper include:(1)Building a knowledge graph of the urban road networkIn this paper,after topological inspection of urban road data and deletion of orphan roads,the road data is abstracted into graph data with the idea of pairwise representation.After completing knowledge extraction of the road data,the graph data is stored in the Neo4 j graph database,and semantic information such as road class and road coordinate points are stored in the graph database as attribute values of graph nodes,completing the construction of an urban road knowledge graph.Afterwards,the urban road selection problem is converted into a graph node classification task in the knowledge graph,and graph neural networks are used to learn the knowledge and historical experience of road selection,providing knowledge support for the construction of an end-to-end road selection network.(2)Selection of model construction and evaluationBased on the basic principle of graph convolutional neural network,the transfer and aggregation of spatial information of graph nodes based on the spatial method of graph convolution is analysed.The urban road selection model is designed by combining Graph SAGE with knowledge graph,and the influence of different parameters in the model on the selection results is investigated.The research results show that the road selection method based on knowledge graph and graph convolutional neural network can effectively extract and aggregate the semantic features,geometric features and spatial features of roads.In comparison with the traditional road selection methods,the selection model proposed in this paper avoids the interference of human factors such as relevant thresholds and weight settings,realizes automatic road selection and improves the degree of automation.In the road selection experiments at scales from 1:250,000 to 1:1 million,the accuracy of the automatic road selection model built in this paper reached 95.8%,which is significantly higher than that of the MLP algorithm 66.9% and the SVM algorithm 84.8%.In the generalisation test of the model,for the road network of Kaifeng,the accuracy of the model selection is 91.08%,and the selected roads can better The road connectivity and density of the road network were maintained.