Research On Parallel Visualization Of Flow Field Based On Streamline Similarity

Flow visualization is one of the focuses in the research of scientific computing visualization.Among the various methods of visualizing flow fields,streamline visualization methods are widely used because they can intuitively show the essence and variation of flow field.However,the streamline visualization method can easily lead to clutter of the visualization effect,which is particularly serious in the three-dimensional flow field.With the improvement of the calculation accuracy,the calculation of the streamline generation process also increases in time,which affects the visualization efficiency.In view of the above problems existing in the streamline visualization method,this thesis studies the parallel visualization method based on streamline similarity.The main work is as follows:Firstly,the streamline generation method is studied,and a streamline generation method with adaptive integration step is implemented,which includes point positioning,velocity vector interpolation,integral method and so on.After the streamlines are generated,the feature points of streamline are extracted by using the spatial properties of streamline curvature and torsion.The similarity of streamline shape is judged by the feature points,and the method of judging the similarity of the distance between the streamlines is adopted to design the method of judging the similarity of three-dimensional streamlines,reduced the clutter of streamlines in visual results.Secondly,aiming at the problem of low computational efficiency of streamline generation algorithm,this thesis uses parallel computing technology to improve computational efficiency.The parallel strategy applied to the streamline generation process is analyzed.Based on the parallel strategy of seed points,a parallel task partitioning method based on task equalization and redundant task redistribution is proposed.This method can allocate tasks to each node relatively evenly and avoids the overload of single node.Aiming at the problem of congestion waiting when judging the similarity in parallel,this thesis proposes a double-buffer queue-based loop detection method to avoid the problem of blocking waiting.Finally,in the Windows environment for experimental analysis and verification.The visualization results based on streamline similarity and the computational efficiency of the algorithm under serial and parallel environment are compared and analyzed.The experimental results show that by judging the similarity between streamlines,the number of streamlines can be reduced to reduce the clutter of visual results.At the same time,the parallel algorithm can significantly reduce the flow visualization time and effectively improve the efficiency of flow visualization.