基于MB栅格回波强度和改进BPNN的底质分类

Seabed sediment classification Based on Multibeam grid echo intensity and improved BP Neural Network

提出了基于多波束栅格图像和改进神经网络的底质分类方法,研究了多波束栅格回波强度的提取方式和改进的反向传播(BP)神经网络.论述了波束脚印包络内以采样数量进行等角度栅格分配获取回波强度所在位置序列值,并在传统BP神经网络基础上附加动量因子和自适应学习率,同时为激活函数添加斜率和偏置可随误差信号进行修正.改进的BP神经网络不仅可以提高神经元的自适应能力,而且可以明显加快算法的收敛速度.利用提出的方法进行底质分类,实验结果表明,提出的方法显著提高了海底底质分类的分辨率和精度.

A method of seabed sediment classification based on multibeam grid image and im- proved BP neural network was proposed. Meanwhile echo intensity data extracted in grid way and improved BP neural network were studied. The locations of echo intensity in multibeam footprint envelope was allocated in equal angle grid way by reference to sampling numbers, and the traditional BP neural network was improved by means of adding momentum factor and a- daptive learning rate,adding slope and offset in the activation function,which were discussed in the paper. The improved B that but also significantly P neural speed up network can not only improve adaptive ability of the neu the co classification was accomplished with it has higher resolution and precision nvergence speed of the algorithm. Then seabed sedi- the proposed method, the experiment results show than traditional classification methods.