| Map generalization is a process to meet the needs of users' cognitive needs and map scale reduction.It is geared to the combination of attribute field and graphic field,which is the process of the graphic presentation and information processing.Along with the development of the network,the transmission efficiency and cost of massive graphics information data have been put on the agenda and how reasonable filter out redundant information which is not related to application to improve the transmission speed and reduce the cost also caused the attention of the cartographic community.POI is one of the most important contents of mobile map and network map,with the increasing number of users' needs,the amount of data is also growing with each passing day.With the development of high performance computing methods such as distributed computing,grid computing and cloud computing,it provides a powerful support for the improvement of map generalization efficiency.In this paper,which is based on the point generalization parallel computing,the network structure types and road grade composition which are driven by different scale are analyzed.Under considering the field force constraints of road network to the POI and load balancing and other related constraint conditions,the effect of data partitioning granularity on the result of POI partition and its effect on the performance of POI generalization parallel computing are explored.This study explores the law of the data partitioning method based on the relationship of the elements,and enriches the theory and method of map generalization parallel computing.The main contents and innovations of this thesis are as the follows:1)Through the re-generalization of point feature generalization operator and combining the point feature own characteristic and the generalization nature,choose the dual operator model,which is composed of content selection and structure simplification,as the core of this paper.Discuss the point generalization parallelizability combined with data correlation,process data size and the extension of parallel program and from the two aspects of task parallel and data parallel.Finally,put forward the viewpoint that the data parallel is more suitable for the point generalization parallel computing.2)From the perspective of space mapping,spatial clustering,spatial index of three types of data partition method,the existing vector data partition methods were compared.And evaluate the Hilbert encoding partitioning algorithm,the N-KD tree partitioning algorithm,the R*tree partitioning algorithm,the K-Means++algorithm and the hierarchical road network partitioning algorithm based on the spatial aggregation,distance between clusters,internal cluster distance,runtime of data partition and load balancing in five aspects.Through the above comparison and analysis,it can be concluded that the hierarchical road network partitioning approach is more appropriate for the POI generalization parallel computing.3)In order to make the hierarchical road network partitioning approach can effectively partition,this paper presented the partition mechanism oriented rules about road mesh building.And put forward the vector data partition oriented scale-network mapping sets combined with the road structure and the scale of the road grade.4)Put forward the Partition Granularity calculation model and the performance prediction method of point feature generalization parallel program taking into the appropriate granularity.Combined with hierarchical scale of road network and the distribution of POI,put forward the Partition Granularity calculation model based on the measuring method of point cluster metric information and the statistical description of point pattern,and through the verification shows that the model can effectively describe the central tendency of the number of POI in road mesh.On the basis,it completes the reliability testing of POI generalization parallel computing performance prediction method and the discussion of optimal data partitioning granularity. |
PDF Full Text Request