Research On Method Of Spatial Index For GIS

In recent years, the Geographic Information System (GIS) gains rapidly progress, which is widely used in public administration, science and business. GIS is now becoming a very active research field in computer science and technology. In GIS the spatial datum volumes, which are trending to large, are transacted by improving technique of spatial index. Due to the complexity of spatial data and high requirement of queries, researching methods of spatial index with high efficiency has become important in field of GIS.Firstly, this paper makes a survey of the recent work about GIS and technique of spatial index, and describes the characters of spatial datum. Then, it makes some comparisons and analyses of spatial index techniques that are proposed in the literature, also discusses methods on improving and designing spatial indexes.For most of spatial indexes have been proposed do not provide proper tradeoffs between efficiency of storage and operations, we present a novel spatial index method, called HSIBGR (Hybrid Spatial Index Based on Wide Grid and Clustering R-tree in GIS), according to principles of division of zone and Object Oriented. The method partitions research zone to some great grids, and establishes the first index. Then, it builds the second index for every grid based on R-tree. It greatly saves storage space and decreases time of index operation.According to the gathered characters of spatial datum, we present a new algorithm, termed as C-Linear (A Linear Splitting Algorithm of R-tree Nodes Based on Clustering Technique of Partition). Then, the C-Linear to be used in R-tree nodes splitting process optimizes organization of spatial datum and improves efficiency in spatial index.Secondly, the paper describes operations of insertion, deletion and searching of HSIBGR with C-Linear in detail.Finally, we employed C++ language to design algorithm of C-linear, R-tree and HSIBGR, and then design various scenarios to do experiments under Linux enviroment. The results show the validity of HSIBGR is better than that of Guttman's R-tree, and the efficiency of C-Linear excels that of quadratic splitting algorithm of R-tree nodes.