Optimization Method For Areal Feature Matching Considering Context-dependent Similarity

With the vigorous development of application based on spatial location service,the requirement of spatial data quality is getting higher and higher,the requirement of data update cycle is getting shorter and shorter in order to keep spatial data verisimilitude.The key technique in improving spatial data quality and data updating is identical entity matching.Areal features,like buildings and habitations,as the most frequently used elements in spatial location service,with the characteristics of dense distribution and complex shape,combined with the objective reasons such as position deviation and multi-scale,have been the most complex and important part in the process of identical entity matching.Through the comprehensive analysis of the domestic and foreign existing areal features matching method.This paper,based on context-dependent similarity of areal feature and the matching strategy of global optimization,habitations as the research object,design1:1 and 1:M areal features automatic matching method to solve the problem of multi-scale areal features matching,the main works and innovations are as follows:(1)Study on method to determine threshold value(matching parameters)automatically.Based on the analysis of the traditional process of setting matching parameters.This paper put forward a method to optimize the matching parameters by selecting a set of matching samples,repeated training and combining the feedback of the matching result.Compared with previous methods for setting matching parameters according to experience,this method is more objective and avoids the uncertainty of the matching caused by human subjectivity to set parameters.(2)Study on method to optimize searching efficiency and accuracy of candidate matching set.The spatial grid index of areal features is built to avoid the global traversal of the data set,meanwhile,using the maximum distance of the same entity and the geometric similarity threshold to determine the final candidate matching set.This method avoids the absence of the same entity while optimizing the efficiency of searching the candidate matching set.(3)Put forward a new method to evaluate context-dependent similarity of areal features.Based on the analysis of existing evaluation methods of context-dependent similarity,this paper propose to take the features that distance nearest areal feature in north,south,east and west as the set of the context-dependent spatial structure of areal feature.This method takes full account of the neighboring relations of areal feature,and realizes the algorithm quickly.(4)Put forward a multi-scale areal features matching method which takes into account the context-dependent similarity and global optimization matching strategy.Through the comprehensive analysis of the influence of position deviation on areal features matching under the condition of multi-scale,it can reduce the influence while taking into account the context-dependent similarity and global optimization.Therefore this paper proposes a multi-scale areal features matching method which takes into account the context-dependent similarity and global optimization matching strategy.This method,under the condition of multi-scale,can effectively overcome the influence of position deviation on matching and greatly improve the matching accuracy,especially for the habitations of high shape homogeneity and position intensive.