基于特征提取和RBF神经网络的ECT流型辨识

Flow pattern identification based on feature extraction and RBF neural network for Electrical Capacitance Tomography

针对传统ECT流型辨识方法效率低的问题,提出了一种基于特征提取和径向基函数神经网络相结合的ECT图像流型辨识的方法,该方法通过对各种特征参数的定义,完成对ECT系统测得的电容值进行特征提取,然后将提取的特征值作为RBF神经网络的输入完成流型辨识。仿真和实验结果表明,与基于BP神经网络的图像流型辨识方法相比,该方法具有识别速度快和效率高等优点,为ECT图像流型识别的研究提供了一个新的思路。

To improve the traditional methods of identification of flow pattern recognition rate,a flow pattern identification method of ECT images based on the feature extraction and radial basis function neural network is presented.In this method, the definition of feature parameters is presented,feature extraction is finished according to the capacitance values measured from ECT system,and the feature values extracted are input to RBF network to finish flow recognition.Experimental results and simulation data indicate that compared with the method of BP neural network,the new method is superior both in speed and in efficiency,and this method presents a new feasible and effective way to research the image flow pattern identification for electrical capacitance tomography system.