A review of artificial intelligence-based seismic first break picking methods

Seismic first break picking plays a crucial role in providing vital information concerning subsurface structures and seismic activities，thereby holding significance for seismic exploration and geological research. The automatic and accurate picking of first⁃arrival waves from low signal⁃to⁃noise ratio data has garnered consi⁃ derable attention from scholars. This paper provides a comprehensive review of artificial intelligence ⁃ based methods employed for seismic picking. It presents an in⁃depth analysis of the principles，characteristics，and de⁃ velopmental trajectory of five distinct types of methods：clustering，support vector machines（SVM），back⁃ propagation neural network（BPNN），convolutional neural networks（CNN），and recurrent neural networks （RNN）. Clustering，SVM and BPNN methods demonstrate a relatively intuitive and interpretable nature，al⁃ beit requiring manual feature extraction. Conversely，CNN and RNN methods possess the ability to autono⁃ mously learn seismic data features，yet they rely on substantial volumes of labeled data to facilitate their learn⁃ ing process. Furthermore，this paper discusses the challenges and future research directions of seismic first break picking. Specifically，it emphasizes the imperative need to further advance the real ⁃ time capabilities for picking first break under extremely low signal⁃ to⁃ noise ratios and to further develop the lightweight of the net⁃ work. © 2024 Editorial office of Oil Geophysical Prospecting. All rights reserved.