Research On Spatial Index Method Based On Mean And Standard Diviation

Along with the development of Geographic Information Systems (GIS), Image Recognition, Computer-Aided Design and Manufacturing (CAD / CAM), Communications Industry, the applications of spatial database technology are more extensive. Spatial index which ranged between spatial objects and spatial operations, is a supplementary spatial data structure. Through its screening, a large number of spatial objects which is not related to specific spatial operations are ruled out, it can reduce operational costs and improve the overall performance of operational costs and improve the overall performance. Therefore, spatial index occupies a very important position in the field of spatial database.In this paper, after the in-depth study of spatial index's theory and related technologies, we focus on two types of mainstream technologies which are Quad-tree family and R-tree family. QR-tree is a hybrid spatial index which is based on quadtree and R-tree. Considering the inadequate of QR-tree such as performance degradation in the frequently updated environment, index establishing algorithm and node splitting algorithm are improved in the new index which is called MVQR-tree in this thesis.Firstly, in the index building phase, QR-tree based on the current value will result in the index update every time a data value changes, so it is affected by huge update overhead. But in the MVQR-tree, every dimension of spatial data items is represented by (mean, variance) coordinate. The MVQR-tree can reduce the number of index update largely and reduce update overhead significantly since the index is adjusted only when the data moves out its interval. The QR-tree based the traditional hierarchical methods (minimization of area enlargement) to split a node which overflows into two new ones brings lots of invalid region and overlapping region in the index space. MVQR-Tree introducing k-means clustering algorithm to split the node by multi-path segmentation can increased the degree of similarity with the group of objects, improve the query speed of the process of pruning and get a better query performance. In the end, simulation experiments are carried out. The experimental results show that the MVQR-Tree has a better overall performance in the environment of spatial objects with frequent updates.