Error-bounded and Number-bounded Approximate Spatial Query for Interactive Visualization

Abstract

AbstractIn the big data era, an enormous amount of spatial and spatiotemporal data are generated every day. However, spatial query result sets that satisfy a query condition are very large, sometimes over hundreds or thousands of terabytes. Interactive visualization of big geospatial data calls for continuous query requests, and large query results prevent visual efficiency. Furthermore, traditional methods based on random sampling or line simplification are not suitable for spatial data visualization with bounded errors and bound vertex numbers. In this paper, we propose a vertex sampling method—the Balanced Douglas Peucker (B-DP) algorithm—to build hierarchical structures, where the order and weights of vertices are preserved in binary trees. Then, we develop query processing algorithms with bounded errors and bounded numbers, where the vertices are retrieved by binary trees’ breadth-first-searching (BFS) with a maximum-error-first (MEF) queue. Finally, we conduct an experimental study with OpenStreetMap (OSM) data to determine the effectiveness of our query method in interactive visualization. The results show that the proposed approach can markedly reduce the query results’ size and maintain high accuracy, and its performance is robust against the data volume.