Edge recognition of magnetic anomaly source body based on convolutional neural networks in Red Sea Basin

The Red Sea Basin is one of the youngest marine basins, experiencing three stages of rift formation, early magmatic activity, and rift expansion. The fault system and uplift pattern are controversial research points. It can provide effective basis for delineating geological units and dividing fault structures by recognizing the edge information of field source bodies with magnetic anomaly data. However, traditional methods for identifying the boundaries of magnetic anomaly source bodies are affected by factors such as the depth of the source body, magnetization direction, and mutual interference between magnetic anomalies, which can lead to errors in subsequent interpretation work. The latest development of convolutional neural networks has strong feature representation and deep learning capabilities. This paper proposes an edge recognition method based on convolutional neural networks. Firstly, a network architecture for identifying the boundaries of magnetic anomaly sources was designed based on the U-Net network. Then, models with different parameters such as location, scale, quantity, and physical properties were selected to construct a large amount of high-quality sample data for training the network. Finally, a model experiment was designed, taking into account the effects of burial depth and tilted magnetization. The effectiveness of the proposed method was verified by comparing it with traditional edge recognition methods. Finally, based on the geological gravity data of the Red Sea and the Gulf of Aden, the division of the Red Sea and Gulf of Aden fault and uplift system was completed.