Research On Parallel Streamline Generation Algorithm Based On Feature Extraction

Flow field visualization is one of the important fields of scientific computing visualization,and streamline visualization is widely used because of its intuitiveness.However,the visualization of streamlines is easy to cause visual clutter effects such as too dense streamlines;and with the increase of the total amount of data and data accuracy,the generation process of streamlines becomes quite time-consuming,which seriously affects the efficiency of visualization.Aiming at the problems existing in the current streamline visualization method,this paper studies a parallel streamline visualization method based on feature extraction.First,a streamline generation method with adaptive integration step size is proposed,which includes information entropy calculation,seed point location,integration method,vector field reconstruction and other processes.Before generating streamlines,obtain the complete information entropy of the entire flow field,set seed points and integration steps according to the size of information entropy in different regions,and use the combination of Euler method and fourth-order Runge-Kutta method to generate flow Lines,after generating streamlines,reconstruct the intermediate vector field and calculate the conditional entropy,and judge whether to continue to generate streamlines according to whether the conditional entropy converges.This method can reduce the clutter in the streamline visualization results and avoid generating unnecessary streamlines.Second,a parallelization method based on the combination of seed points and data blocks is proposed.In order to solve the problem of low efficiency of streamline calculation,this paper adopts parallel computing technology to improve the efficiency of streamline calculation.When generating streamlines,the seed point set is divided according to the number of seed points.This method assigns tasks with similar numbers to each node to ensure that the tasks of each node are equal;in the process of calculating information entropy,reconstructing intermediate vector field and calculating conditional entropy,the entire flow field is divided into data blocks,so as to avoid the situation that the load of a single node is too large.Finally,the experimental comparative analysis and verification are carried out.The visualization results and algorithm efficiency of the parallel streamline generation algorithm based on feature extraction and the traditional algorithm are compared and analyzed.Experimental results show that the arrangement of seed points guided by complete information entropy can reduce the clutter of visualization results;at the same time,the parallel generation of adaptive streamlines and algorithms can significantly reduce the time of flow field visualization and improve the visualization efficiency.