Abstract: The present disclosure relates to a seismic imaging free gas structure identification method and system, The method includes: acquiring a seismic imaging dataset, and annotating free gas structures in the seismic imaging sample data set to obtain a training dataset; using a generative adversarial network (GAN) to expand the training dataset for network training, wherein the GAN includes a generative network and a discrimination network; conducting domain conversion on original label data of the free gas training samples to obtain annotating labels: training a Bayesian neural network according to the free gas training datasets and labels, to obtain a free gas structure identification model; acquiring actual seismic imaging data of a target work area; and conducting, according to the actual seismic imaging data, identification by using the free gas structure identification model.

Abstract: The present disclosure relates to a seismic imaging free gas structure identification method and system. The method includes: acquiring a seismic imaging dataset, and annotating free gas structures in the seismic imaging sample data set to obtain a training dataset; using a generative adversarial network (GAN) to expand the training dataset for network training, wherein the GAN includes a generative network and a discrimination network; conducting domain conversion on original label data of the free gas training samples to obtain annotating labels; training a Bayesian neural network according to the free gas training datasets and labels, to obtain a free gas structure identification model; acquiring actual seismic imaging data of a target work area; and conducting, according to the actual seismic imaging data, identification by using the free gas structure identification model.