基于迁移学习，设计一套岩石边坡微地震事件检测算法流程，用于自动化处理岩石边坡数据。基于海量人工标注的天然地震数据进行训练，得到深度学习预训练模型，并利用少量人工标注的微地震数据进行微调，使得模型可以适用于滑坡体微地震数据。采用实际标注数据进行测试，结果表明，基于迁移学习模型的查准率和查全率分别可达0.884和0.91。分析认为，在迁移学习流程中，深度学习模型减少了对于标注数据的依赖，同时可以仅经少量迭代即可得到鲁棒的、高精度结果。该模型部分程序是开源的，可以将其迁移到更多区域的微地震事件检测工作中。
In this article, we introduce a transfer learning-based landslide microseismicity detection model, which can automatically pick up microseismicity occurring on the slopes in more accurate means. The deep learning model is first trained using a huge amount of manually labeled seismic events to obtain a well pre-trained model, then, the pre-trained model is fine-tuned by a small number of manually labeled microseismic events that have occurred on the slope to account for landslide microseismicity detection. The results suggest that our model achieves a rate of 0.884 and 0.91 in recall and precision test using unknown events that occurred on the slope, respectively. The proposed transfer learning-based training procedure not only significantly reduces the demand on the labeled training data on the slope, but also achieves a more robust and accurate model using a small number of integrations when applied to slopes. We open source the main function of the model, which can also be applied to other slopes.
2024,45(2): 20-27 收稿日期：2023-12-04
DOI：10.3969/j.issn.1003-3246.2024.02.003
基金项目：基于深度神经网络的体波面波联合反演算法，中央级公益性科研院所基本科研业务费（项目编号：DQJB23R31）
作者简介：蔡育埼（1999—），男，硕士研究生，研究方向：地震信号检测。E-mail：caiyuqiming@foxmail.com
*通讯作者：于子叶（1989—）,男，副研究员，博士，研究方向：地震信号处理。E-mail：yuziye@cea-igp.ac.cn
参考文献：
刘泽民，李俊，苏金蓉，等. 2022年四川泸定MS 6.8地震余震序列的自动构建与地震活动性分析[J]. 地球物理学报，2023，66(5)：1 976-1 990.
Allen R V. Automatic earthquake recognition and timing from single traces[J]. Bulletin of the seismological society of America, 1978, 68(5): 1 521-1 532.
Baer M, Kradolfer U. An automatic phase picker for local and teleseismic events[J]. Bulletin of the Seismological Society of America, 1987, 77(4): 1 437-1 445.
Devlin J, Chang M W, Lee K, et al. Bert: Pre-training of deep bidirectional transformers for language understanding [C]// Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies Volume 1 C Long and short Papers). Minneapolis, Minnesota: Association for Computational Linguistiscs, 2018.
Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Columbus, OH, USA: IEEE, 2014: 580-587.
Hu J P, Yu Z Y, Kuang W H, et al. Application of machine learning methods in arrival time picking of P waves from reservoir earthquakes[J]. Earthq. Res China, 2020, 34(3): 343-357.
Mostafa Mousavi S, Sheng Y, Zhu W Q, et al. STanford EArthquake Dataset (STEAD): A global data set of seismic signals for AI[J]. IEEE Access, 2019, 7: 179 464-179 476.
Mostafa Mousavi S, Ellsworth W L, Zhu W Q, et al. Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking[J]. Nature communications, 2020, 11(1): 3 952.
Whiteley J S, Chambers J E, Uhlemann S, et al. Geophysical monitoring of moisture induced landslides: a review[J]. Reviews of Geophysics, 2019, 57(1): 106-145.
Yu Z Y, Wang W T. Lppn: A lightweight network for fast phase picking[J]. Seismological Research Letters, 2022, 93(5): 2 834-2 846.
Yu Z Y, Wang W T, Chen Y N. Benchmark on the accuracy and efficiency of several neural network based phase pickers using datasets from China Seismic Network[J]. Earthquake Science, 2023, 36(2): 113-131.
Zhu J, Li Z F, Fang L H. USTC-Pickers: a Unified Set of seismic phase pickers Transfer learned for China[J]. Earthquake Science, 2023, 36(2): 95-112.
Zhu W Q, Beroza G C. PhaseNet: A deep-neural-network-based seismic arrival-time picking method[J]. Geophysical Journal International, 2018, 216(1): 261-273.
