Study On Implementation Of Fast Visualization Of Geographic Vector Data

With the widely appliance of massive detectors like remote sensors and aerial surveyors, countries all over the world are attaining data of Multi-source Aero Spatial Image Information that are increasing in a geometrical progression. People have arrived the big-data times. Then,how to visualize those big data becomes a focus problem for the geographical information system.By so far, the professional software of geographical information system mostly are desktop software which are installed on the personal computer, the data are stored in the local storage and all of the compute works are done by this computer. So, it is unable to deal with visualization of the large scale data. To solve this increasingly prominent problem, we proposed a fast visualization method based on cluster server. This thesis launches the research from the following aspects.Firstly, we designed a framework of visualization using B/S structure based on high performance computer. Users access the web page and then can browse the map. The maps are divided into tiles. The middle part is the web server. It is the bridge of the front part and the cluster server. The well visualized tiles are cached on the cluster server file system and are managed by the web server. Moreover, the web server balances the workload of the cluster server. Finally is the core part of the framework, the visualization server. It is buildup by the visualization engines and spatial database on several cluster nodes. This frame eases the burden of the personal computer by taking the advantage of the cluster server.Secondly, in order to reduce the rendering time of one tile which may still has large numbers of features to render, we proposed a multi-thread parallel visualization method on multi-core server based on the traditional visualization method. As the features of the geographic data are irrelevant and are rendered one by one. We can partition the features into blocks, and then one thread renders one block of features. All of the threads render the features individually with no mutual communication.Finally, although it will be speed up through parallel visualization, but how to visualize efficiently is another problem to be considered. We proposed a continuous data dividing method and a discontinuous data dividing method which are based on geographic data parallel visualization. It is testified by experiment that the load balancing efficiency of the continuous divide data method reaches 85% and discontinuous divide data method reaches 97%.