基于模糊神经网络的涡结构识别方法研究

Vortex Structure Recognition Using Fuzzy Neural Network

研究气动光学传输效应产生的机理是红外成像末制导的共性基础技术之一,基于涡结构对光学传输效应进行建模是一种非常有效的方法,而涡结构的识别是其必要前提。文中提出一种新的涡结构识别方法,把折射率场经小波变换后的系数矩阵等效为具有一定纹理结构的图像,计算图像的共生矩阵及其统计量,由于涡结构模式复杂,特征量较多,设计了等价结构的模糊神经网络进行涡结构识别。与小波分解后直接提取特征量的识别方法相比,本文的方法从空、频角度更加准确全面地表征湍流涡结构模式,计算机仿真结果表明该方法优于神经网络的识别效率。

Investigation on the principle of the aero-optic propagation is a fundamental technology for infrared imaging homing guidance. It is an effective approach to study the model of the aero-optic propagation based on the vortex structure, and the vortex structures recognition is its necessary premise. A new vortex structure recognition method is proposed in this paper. Firstly, wavelet transform is applied to the refractive index of the flow field and wavelet module matrix is obtained. Then, the matrix is transformed to an equivalent image containing texture information. And the co-occurrence matrix and statistical parameters of the image are calculated. Furthermore, a fuzzy neural network with equivalent structure is applied to accomplish the recognition in consideration of the complex vortex structures and more characteristics. Compared with direct recognition after wavelet decomposition, the proposed method can express more characters of turbulence vortex structure in space-frequency domain. Simulation results demonstrate the validity of this method and its better efficiency than that of traditional NN.