Flow Feature Detection Based On Entropy And Clifford Algebra

Flow visualization is an important research direction in scientific visualization. The flow feature detection method can extract the most important or interested area of the flow field. The traditional flow feature visualization method based on pattern match cannot use the full vector field information and then cannot detect flow feature accurately. The patten match method based on Clifford algebra can improve the accuracy, but followed by a huge amount of computation. Therefore, a feature detect method is proposed on the basis of information entropy and Clifford algebra. This method computes entropy and chooses high entropy regions to do pattern matching. It can improve the computing efficiency. In addition, the contents we concerned is different when the type of flow field is different. Standard templates are hardly to cover all kinds of pattern structure. So this paper proposed a flow feature detection method based on customed template. The method allows user to select interested pattern and then uses this pattern to detect other similar regions in the flow field. The main work accomplished in this paper is as the following:(1) A feature detection method based on Clifford algebra is implemented. The feature templates are set up according to the critical point type of the flow field. Then the flow field is matched with the feature templates by Clifford convolution. The flow feature points and types can be identified according to the convolution result.(2) A feature detection method is proposed on the basis of information entropy and Clifford algebra to improve the calculation efficiency of Clifford algebra. This method used Clifford algebra to do patten match on high entropy area. It avoids a large number of invalid template matching and improves feature detection efficiency.(3) A method based on customed template is proposed to detect feature of flow field. This method lets users customize the template based on their own needs. It extends the adaptive capacity and scope of the above mentioned method.(4) Software tools are developed to visualize 2D and 3D flow fields according to the methods proposed in this paper.