A NOVEL APPROACH OF INDEXING AND RETRIEVING SPATIAL POLYGONS FOR EFFICIENT SPATIAL REGION QUERIES

Abstract. Spatial region queries are more and more widely used in web-based applications. Mechanisms to provide efficient query processing over geospatial data are essential. However, due to the massive geospatial data volume, heavy geometric computation, and high access concurrency, it is difficult to get response in real time. Spatial indexes are usually used in this situation. In this paper, based on k-d tree, we introduce a distributed KD-Tree (DKD-Tree) suitbable for polygon data, and a two-step query algorithm. The spatial index construction is recursive and iterative, and the query is an in memory process. Both the index and query methods can be processed in parallel, and are implemented based on HDFS, Spark and Redis. Experiments on a large volume of Remote Sensing images metadata have been carried out, and the advantages of our method are investigated by comparing with spatial region queries executed on PostgreSQL and PostGIS. Results show that our approach not only greatly improves the efficiency of spatial region query, but also has good scalability, Moreover, the two-step spatial range query algorithm can also save cluster resources to support a large number of concurrent queries. Therefore, this method is very useful when building large geographic information systems.