Geographic Traffic Information Based Fast Edge Bundling Visualization Research

As one of the most important ways to reflect the correlation of data, graph has beenwidely applied in different ways. Edge, which serves as a visual primitive used in therelated data in code pattern, plays an important role in graph visualization. However, withan ever increasing data volume, tens of thousands of edges shelter each other during thevisualization process, which will not only damage the overall artistic effects ofvisualization, but will also lead to vision confusion where users cannot draw informationneeded from the graph. Several edge bundling technologies emerge as the time requires,which lower vision complexity by altering edge layout. Considering the status quo ofseveral edge bundling algorithms and their failure to combine visualization withgeographic information, this paper puts forward a new edge bundling method which isbased on geographic traffic information. And when combined with heat map, the amountof information conveyed in a graph can be greatly increased and reduced the visual cluttereffectively.First, an in-depth research and analysis of several classic edge bundling algorithmsshow that traditional edge bundling methods only focus on visualization of the graph itselfwithout an effective combination with the geographic information, thus giving rise to amethod which combines graph visualization with geographic traffic informationvisualization. In this algorithm, the shortest traffic route can be worked out with anadvanced Dijkstra algorithm and an introduction of the existing traffic information into theedge bundling process. On this basis, edges in the original graph are replaced by Beziercurve while a series of consolidation strategies are set in order to further enhance theedges’ degree of polymerization.Second, taking edge bundling based on geographic traffic information as a basis, heatmap visualization is used. Graph data is expressed by graphicalization with a use of colors.Interpolating points generated during visualization are taken as hot spots, whose values ofinfluence on the regions are calculated to be the final heat value by Gaussian kernelfunction. Faster operation can be achieved by the use of GPU simultaneousmultithreading.Besides, areas in the original graph can be segmented by quadtree structure. Makinguse of the distribution characteristics of the Gaussian function, hot spots can be hookedwith panel points according to the domain of influence by the retrieval of quadtree. Whencalculating the heat value, less problem scale and calculation time complexity can beachieved by the screen of hotspots needed calculating by the retrieval of quadtree,according to the geographic information represented by the heat value.In the end, the final visualization effects are analyzed by the comparison withforce-directed edge bundling algorithm and geometry-based edge bundling algorithmthrough the experiments. And the layers of the quadtree’s influence on the timecomponents are analyzed by the time complexity under different layers when adjusting thelayers of the quadtree. By this means, the validity of this algorithm is proved.