基于图像小波包信息熵和遗传神经网络的气-液两相流流型识别

Identification Method of Gas-Liquid Two-Phase Flow Regime Based on Image Wavelet Packet Information Entropy and Genetic Neural Network

根据小波包变换能够将图像信号按不同尺度进行分解的特性,提出了基于图像小波包信息熵特征和遗传神经网络相结合的气-液两相流流型识别的新方法。该方法采用高速摄影系统获取水平管道内气-液两相流的流动图像,经过处理,对图像进行多分辨率分析,提取小波包变换系数的信息熵特征,用主成分分析法降低特征维数构成特征矢量,作为流型样本对遗传神经网络进行训练,实现了对流动图像的流型智能化识别。结果表明:图像小波包信息熵特征可以很好地反映各流型之间的差异;遗传神经网络结合遗传算法和BP算法各自优点,具有收敛速度快、不易陷入局部极小的特性,网络识别率为100%。

Based on the characteristic that wavelet packet transform image can be decomposed by different scales, a flow regime identification method based on image wavelet packet information entropy feature and genetic neural network was proposed. Gas-liquid two-phase flow images were captured by digital high speed video systems in horizontal pipe. The information entropy feature from transformation coefficients were extracted using image processing techniques and multi-resolution analysis. The genetic neural network was trained using those eigenvectors, which was reduced by the principal component analysis, as flow regime samples, and the flow regime intelligent identification was realized. The test result showed that image wavelet packet information entropy feature could excellently reflect the difference between seven typical flow regimes, and the genetic neural network with genetic algorithm and BP .algorithm merits were with the characteristics of fast convergence for simulation and avoidance of local minimum. The recognition possibility of the network could reach up to about 100%, and a new and effective method was presented for on-line flow regime.