提出了一种应用数字图像处理技术对模拟尾流气泡幕分类识别的新方法.文章介绍了BP神经网络的基本结构及其工作原理，通过仿真测试了BP神经网络对模拟尾流气泡幕图像的模式分类.应用灰度图像统计矩法得到了均值﹑归一化系数﹑三阶矩﹑一致性和熵等特征量，设定神经网络学习率为0.1时经过14次循环可以达到训练目标误差为0.001，此时网络对不同压强下的尾流气泡幕分类正确率到达100%.这种方法在处理尾流图像时具有直观、高效、精确等特点，易于应用于对尾流探测、识别等工程技术中.

A new method which classification of simulated wake bubble films（SWBF）can be obtained by using imaging processing methods. The paper introducts the basic structure and working principals of the BP neural network, the simulation tested classification of SWBF image based on BP neural network. The characteristic quantities such as mean value,normalized coefficient,the third moment,uniformity, entropy can be extracted based on gray histogram statistical moment. After 14 epochs, training error can be reached 0.001 when we setting neural network learning rate is 0.1, while the classification accuracy can be up to 100% under different pressures on SWBF.The method has characteristics such as visibility, high efficiency and accuracy, and can be apt to applied in engineering projects for wakes′ detection and recognition.