Research Of Multi-level Adaptive Indexing Method Of Trajectory Data On Cloud Platform

With the development of satellite positioning technology and the popularization of mobile communication equipment,massive trajectory data generated in human daily life.The storage and indexing of trajectory data is a key research area in urban construction and traffic management.The traditional centralized indexing method is based on single node.There are limitations in dealing with high concurrent read-write and extensibility of massive data.With the development of cloud computing,a series of distributed indexing methods have been proposed,which make up for the shortcomings of traditional methods.However,most of these methods are only suitable for historical static data,which is not suitable for dynamic data.Moreover,the indexing strategy is mainly designed for spatial-temporal range queries,cannot provide good support for various query types.In order to improve the scalability and query efficiency of storage and indexing,this paper proposes a multi-level adaptive trajectory data storage and indexing scheme based on cloud computing platform.First,for the simpleness problem of distributed indexing methed,construct the first-level index based on object identifier and time attributes contained in the trajectory data;Secondly,in order to solve the problem of performance degradation of traditional indexing methods when processing massive high-dimensional data,construct the Hilbert spatial index on the spatial attributes in second-level index,reduce the two-dimensional spatial attributes,and use the balanced partition method of inode and the longest common prefix naming method to store the trajectory data to HBase(Hadoop Database)distributed column database;Finally,according to the spatial distribution of trajectory data,select latitude and longitude dynamically and self-adjust the trajectory data.The trajectory data is gradually ordered with the execution of the query,so that the query performance of the spatio-temporal range is gradually improved.The main contributions of this paper are: 1: In multi-level trajectory index,the first-level index is constructed according to object identifier and time attribute of the trajectory data,the index and storage file are constructed on HDFS(Hadoop Distributed File System)distributed file system,which is suitable for the object identifier and time range query.2: When constructing the second-level spatial index,index nodes constructed by the Hilbert spatial index adopt a balanced division method and the longest common prefix naming method to effectively prune the two-dimensional space,when the query locates a sparse area,there are few index nodes accessed,which improves the efficiency of the spatio-temporal range query.3: For the trajectory data stored in HBase,according to the distribution of trajectory data in space,selected the latitude and longitude dynamically,and then the boundary values in the spatio-temporal range query are used for self-adjustment,so that the trajectory data tends to be ordered as the query is executed,which improves efficiency and performance of spatio-temporal range query,the feasibility and effectiveness of the proposed method are verified by experiments.