中文摘要：

传统的基于拓扑分析方法的特征可视化系统的扩展性、通用性和交互性较差。本文分析了流场的特征，在此基础上提出了一种基于BP神经网络的可选择智能流场特征提取方法，设计了一种三层BP神经网络结构，用户可以对感兴趣的新特征进行选取并进行训练和提取，而无须修改程序。该方法利用神经网络较强的非线性映射能力，提高了系统的扩展性、通用性和交互性。基于上述方法，设计并实现了一个流场可视化原型系统。实验表明，该方法对流场任意特征具有高识别率和较低的误警率、漏报率。

英文摘要：

Feature-based visualization of flow is an important field of scientific visualization. The traditional visualization system based on topology analysis of the flow field does not have scalability, generality and good interaction. Based on the flow feature analysis, this paper presents a selective and intelligent flow feature extraction method. We design a three-lay- ered BP neural network, and the user can select the new feature region they are interested iru Using the strong non-linear ability of the neural network, we are successful in improving the system＇s scalability, generality and interaction. Finally, we introduce a demonstration system based on the above methods, and the test shows that this method has a high recognition rate and a low error-calling rate.