Research On Automatic Matching And Change Detection Of Vector Road Data

Road data is an important part of the basic geographic database, meanwhile, being an vital content to special data such as navigation app, disaster relief, logistics transportation,it's actuality will directly determines the accuracy and efficiency. To promote the economic development, the government has invested a lot of money in infrastructure, especially the building of various road network. In recent years, the road network grows vigorously and change quickly. Currently, one of the most important ways to update the spatial database is incremental updating, which plays an important role in keeping database up-to-date. The incremental updating of spatial data consists of four important processes:(1) Data pre-processing to eliminate topologic ambiguities in the datasets and make sure that the attribute fields are the same between new data and history data;(2) road matching to find same roads in the two datasets; (3) change detection to find the roads have changed. (4) updating the database. The incremental updating has two key processes: road matching, change detection. Change detection can only happen when road matching is finished, and incremental updating can be carried out only when the detection and extraction of change elements is accomplished.In view of the above requirements, the thesis, based on the radial basis function network(RBFN) and the decision tree, focuses on the research of road matching, road change detection and change classification. The main accomplishment is divided into four aspects as follows:Firstly, by concluding and analyzing the current research situation at home and abroad of automatic matching of road network elements and change detection, the thesis further discusses some existent problems in these two aspects on the basis of summarizing the current research methods. Besides, it sums up eigenfactor and type of road change according to the changing rule of road network, which lays a theoretical basis for automatic matching and change detection of road network.Secondly, the thesis proposes a road network matching algorithm considering multiple geometric characteristics with radial basis function network (RBFN). It utilizes length similarity,orientation similarity,shape similarity,distance similarity and topological similarity (topological similarity is the node-degree similarity) and other three spatial feature similarity as the evaluation factors to judge whether the two road arcs are the same or not. In order to solve the similarity indicators' weight allocation in matching, the thesis makes use of RBFN to improve the classic radial basis function.Improved RBFN gives consideration to different road network similarity indicators'different functions in road matching, which makes the radial basis function has anisotropic characteristics. In addition, to achieve the reliable results in road matching,it introduces sigmoid function in the neural network output layer to carry out normalization. Comparing with the BP neural network'effects in road matching,the result shows that RBFN worked more effectively and accurately in sample training and road matching.Thirdly, the thesis proposes a method for road network change detection and classification with decision tree. Setting up the five road change factors such as length,shape, orientation, node-degree and attribute as the character of decision tree, the thesis improves the traditional way which is based on the information gain to create decision tree,calculates quickly the sample data's influence in road change under the influence of the five factors, and the computation will determine the character of decision tree so as to create decision tree including road change classification. In terms of the above steps,the change detection and change classification of road network has been finished.Fourthly, it designs and develops a prototype system of road network change detection and classification based on the theoretical research of this thesis. This system can realize the function of road network data management, simple data pre-processing,road matching,road change detection,change type classification as well as road data query of changing information, attribute, space.