In marine exploration, velocity modeling of complex deep-water salt is an extremely challenging task. On the one hand, complex salt structures exist in a unique environment where seawater coexists with elastic media; on the other hand, the strong amplitude converted waves generated by high-velocity complex salt can affect the accuracy of traditional acoustic velocity modeling. Therefore, it is essential to fully consider these critical factors in order to carry out P- and S-wave velocity waveform inversion modeling for such special structures. Addressing the challenges faced in velocity modeling of complex deep-water salt using towed streamer data, this paper employs a method of boundary coupling between acoustic and elastic wave equations to construct an acoustic-elastic coupled equation (AECE). This equation is more suitable for simulating the propagation process of seismic waves in the overlying seawater elastic media. Through numerical simulations of a three-layered model, we have verified the feasibility of using towed streamer data for S-wave velocity inversion. Subsequently, we derived the waveform inversion gradient operator applicable to the acoustic-elastic coupled equation and established a method for P- and S-wave velocity waveform inversion of complex salt based on towed streamer data. Finally, we validated the applicability and accuracy of this new method through inversion tests conducted on a complex salt media model constructed based on actual overseas geological conditions, as well as through inversion trials using towed streamer data.

Accurate observation system information is the foundation of high-precision seismic data processing. Due to the inability to use conventional GPS positioning for OBN receiver points, specialized technical means are usually required to determine the coordinates of the receiver points. Common secondary positioning methods for receiver points include least squares method, approximate tetrahedron method, search method, fitting surface method, equivalent velocity method, vector synthesis method, etc. However, they are all based on direct wave travel time for positioning. Shallow water OBN data only has direct waves within a small offset distance, making those methods difficult to apply. Another commonly used secondary positioning method is refracted wave secondary positioning, which has low positioning accuracy due to its inability to handle speed gradients, interface bending, and other situations. In response to the problem of inaccurate positioning of receiver points in shallow water environments, this paper proposes a positioning method based on first arrival travel time fitting. Firstly, the first arrival time is accurately picked up from the common receiver point gather. Then, a polynomial is used to fit the relationship curve between the first arrival time and offset. The coordinates of the receiver points are adjusted by the difference between the actual picked travel time and the corresponding value on the curve, and multiple iterations are carried out until convergence. Finally, based on the water depth values measured in the survey, the corresponding Z coordinates can be obtained by projecting the X and Y coordinates of the receiver points, thus obtaining accurate three-dimensional coordinates of the receiver points. The method proposed in this article has been applied to both simulated and actual OBN data, achieving ideal results.

Deep Learning (DL)-based Ground Penetrating Radar (GPR) target recognition methods have been widely applied in geological exploration and infrastructure inspection. However, existing approaches face three main limitations: (1) Most GPR systems operate in single-polarization mode, leading to incomplete acquisition of target scattering information; (2) Traditional deep learning methods risk misclassification when handling B-scan images with similar hyperbolic features from different targets; (3) Direct input of 2D B-scan images into convolutional neural networks incurs high computational overhead. To address these challenges, this paper proposes a multi-polarimetric decomposition fusion method for GPR target recognition based on a lightweight MobileNetV3 network. The proposed method first acquires full-polarimetric GPR data (HH, VH, and VV polarizations) of subsurface targets. Subsequently, H-Alpha decomposition, Freeman decomposition, and Pauli decomposition are performed to extract eight polarimetric parameters characterizing the targets. These parameters are fused into an eight-dimensional feature matrix, which is then fed into a modified MobileNetV3 network integrated with a Squeeze-and-Excitation (SE) attention module for target identification. To verify the effectiveness of the method, four typical targets were classified in the experiments, and the results indicate that using the eight-dimensional feature matrix as the network input can enhance the target classification accuracy. The target classification accuracy can be further improved by incorporating the SE module into the network. Furthermore, compared to conventional ResNet18 and VGG16 networks, the improved MobileNetV3 achieves the highest recognition accuracy (98.75%) while significantly reducing parameter number and model size. The experimental results demonstrate that using an eight-dimensional feature matrix that includes target polarization information as the network input not only provides richer target information but also effectively reduces the redundant information of the input network. This improvement enhances the target classification accuracy while decreasing the matrix size of the input network. Additionally, the lightweight backbone network based on MobileNetV3, integrated with the SE attention mechanism, enhances critical feature extraction capabilities and strengthens discriminative power for target classification. The paper effectively addresses the challenges of insufficient feature discrimination and high computational load in GPR target classification.