智能流场特征抽取与多分辨率可视化

Intelligent Flow Feature Extraction and Multi-Resolution Visualization

分析了流场特征,提出一种基于BP神经网络的可选择智能流场特征提取方法,由用户选择关注特征区域并将该区域作为输入样本进行训练,利用训练后的神经网络对新数据进行识别预处理,抽取出用户关注的特征区域;提出一种基于"鱼眼视图"技术的多分辨率流场特征绘制方法,将原始数据场采用层次细节二叉树表示,以减少绘制数据量.基于上述方法,设计并实现了一个原型系统,对用户关注区域进行详细信息显示,同时保持了整个数据场的概貌,实现了具有良好交互性的可选择流场特征可视化.

This paper presents a selective and intelligent flow feature extraction method based on backward propagation （BP） neural network. By this method, users can select certain regions in concern as input samples to train the BP network. After the training process complete, the BP network is used to preprocess the new field data, to extract the regions in interests. A multi-resolution rendering method is designed based on ＂fisheye views＂ technology, with the field dataset represented as a binary tree structure, to promote the efficiency of the data rendering. Besides, a prototype demonstration system has been also designed, which could maintain the general picture of the data field with high resolution details shown on the interested regions. The result shows excellent interaction with users by our method.