利用局域滤波方法反演地声参数

Geoacoustic inversion using local filter

地声参数的声学反演方法一直是水声学研究领域的热点。针对现有地声参数反演方法在分批分时处理海量数据时不能同时兼顾反演速度和精度、难以满足实际需要的不足,本文基于一个新构造的阵列观测数据向量,提出了一种反演地声参数的局域滤波新方法。该方法首先由前向声传播模型产生一个拷贝场样本集,然后采用加权最小二乘滤波器在阵列实测数据的局域空间内近似地估计了地声参数反演算子(即从数据空间到参数空间之间的映射关系)。本文方法能够充分地利用声场观测信息、避免了整体平滑效应、且海量数据处理时仅需要较少的前向模型计算,具有反演速度快、精度高、兼顾二者的优点,特别适用于对海量数据的分批分时处理及要求实时反演的应用场合。最后,典型浅海环境下的数值仿真和某次地中海实验数据分析结果验证了本文方法的有效性。

Geoacoustic inversion is an interesting topic in underwater acoustics community in recent years. Traditional geoacoustic inversion methods mainly include global-search method and neutral network inversion method. However, the former cannot meet the real-time requirement in practical applications. And the disadvantage of the latter is that the accuracy of the inversion result might not be adequate in some cases. To solve these problems, a geoacoustic inversion method based on the local filter is proposed in this paper to estimate geoacoustic parameters from the measurement of the acoustic field. A new observed data vector including both amplitude and phase information is defined, and then a local weighted least square filter is used to approximate the inverse function which links the geoacoustic parameters with the measured field based on the simulated training pairs generated with an acoustic propagation model. The main advantage of the proposed method is that, it is capable of inverting a few thousand data sets collected at different sites or at different time with a reasonable accuracy, and requires less computation time and could be performed in real time. Finally, the robustness and effectiveness of our proposed method has been illustrated through the numerical simulation and the Mediterranean experimental data.