Research On Spatial Index Technology For Spatial Database

The technique of spatial index is an important research contents. The performance of index will directly influence the performance of database. The search and query are most frequent operations. To improving the speed of query, it needs to build efficient spatial index structure to support the correlative operations. During the several years, the spatial index has come into being an integrated system. Because of the massive, complex and various characters of spatial data, it requires a more effective index structure. The research on spatial index is arousing more and more interests and attentions. Building an efficient index strcture and proposing efficient query methods is becoming important research field. The paper focuses on the creating of spatial index and the query of spatial data based on index, some methods are proposed to solve the above problem.The existing index method can not keep the spatial data relativity in mapping better. The neighbor data can't be stored in the neighbor nodes of index. The bulk-loading method is introduced to preprocessing the data. On the comparison and analysis the index factors, a new bulk-loading method is proposed in this paper for the static data with few changes in database. The method can reduce the overlap area of index. It uses experiment to prove the proposed method. The experiments results show that this method can achieve a higher spatial utilization and reduce the time consumed. The performance of index improved than the previous method.The spatial index structure should be adjusted with the change of data. The essence of dynamic index creating is clustering matter. Through the analysis of existed clustering methods, clustering algorithm based on gird and density is introduced to cluster the data. The lack of existed clustering method is improved. The data is divided into different cluster. It adopts two-level index to create the whole index. Each cluster has its respective index strcture. It used the global R-tree to build the full index. The experiments results show that the proposed method can improve the time and space complexity. According to the existed query processing based on directional relationship query model it can'tachieve better filter results in filter phase. The time complexity is high in the refine step. On the basis of the definition of object's MBR, an intermediate process is inserted in the middle of the filter and refinement step. The directional relationship of target object and reference object can be judged by the reference object falling in the different segmentation. All the probability combination cases are analysised and give the solution method. The improved method could decrease the size of candidate set in the filter step, and thus it can reduce the workload of the refine step and improve the speed of directional relationship. The validity of method is be proved by instance analysis and improve the performance of index. And then, it uses experiments to prove the validity. The experiments results show that the proposed method performs well with respect to both I/O-and CPU- time.The reference object is simplified to a point in nearest neighbor query. The query results have effect in a certain extent. The existed knn method can't well handle the near neighbour query among objects. A k neighbor query method based on the equidistance is proposed. The more exact boundary value and definition of objects are presented. The propose method reduce the workload of the real distance of objects. The validity of method is be proved by instance analysis. The analysis results show that the algorithm achieve better filter effect. The method improved the k neighbor query efficiency and performs well with respect to both in time and space.