利用GA-NN模型反演声速剖面的众源水深数据声速改正

针对当前众源水深数据后处理过程中缺少高精度的实测声速剖面，导致测深数据质量偏低的现状，提出了一种基于遗传算法优化反向传播神经网络（genetic algorithm-back propagation neural network,GA-NN）模型反演声速剖面的声速改正方法。首先，利用历史声速剖面群进行正交经验函数分析，提取特征向量与重构系数范围；然后，结合海区的历史声速场数据训练GA-NN模型；最后，将海表声速数据输入模型反演声速剖面，并分析不同方法下的声速剖面分别进行声速改正后的水深和位置误差。实验结果表明，在复杂的海底地形下，与现有方法相比，所提方法反演的声速剖面更适用于众源水深数据的声速改正，削弱了声速误差的影响，提高了众源水深数据的处理精度。

Correction for Crowd Sourced Bathymetry Data Using GA-NN Model to Inverse Sound Velocity Profiles

  Objectives  Currently , the measured sound velocity profile with high accuracy is lack in the processing of the crowd sourced bathymetry(CSB) data, so that the quality of the sounding data is low. Aiming at this situation above, a method for utilizing the inversion sound velocity profile obtained from the BP neural network model optimized by the genetic algorithm (GA-NN model) is proposed to correct the CSB data.  Methods  Firstly, the eigenvector and the reconstruction coefficient range are extracted from the results via empirical orthogonal functions analysis of the historical sound velocity profile group. Secondly, GA-NN model is trained by utilizing the information of the historical sound velocity profile field. Finally, the surface sound velocity is inputted into the model for getting the inversion sound velocity profile, and statistically analyzed the depth and position errors after correcting the data via different sound velocity profiles.  Results  The experimental results show that in the sea area with complex seafloor topography, the error index of sound velocity profile inverted by the proposed method is smaller, the corrected seabed topography is more fitting to the simulated seabed, and the error of water depth and position is smaller.  Conclusions  The sound velocity profile inverted by the proposed method is more suitable for the sound speed correction in the CSB work, weaks the influence of sound velocity error on seafloor topography, and improves the precision of CSB data.