PSO-based ViT-Seismic: A Vision Transformer Approach for Gas Detection in Seismic Images

Seismic reflection is one of the geophysical methods most used in the oil and gas (O&G) industry for hydrocarbon prospecting. In particular, for some Brazilian onshore fields, such a method has been used for estimating the location and volume of gas accumulations. However, the analysis and interpretation of seismic data is time-consuming due to the large amount of information and the noisy nature of the acquisitions. In order to help geoscientists in those tasks, computational tools based on machine learning have been proposed considering Direct Hydrocarbon Indicators (DHIs). In this study, we present a methodology for detection of gas accumulations based on vision Transformer neural network (ViT) and Particle Swarm Optimization (PSO) scheme. In the best scenario, the proposed method achieved a sensitivity of 75.14%, a specificity of 96.14% and an accuracy of 95.60%. We present some tests performed on Parnaíba Basin which demonstrate that the proposed method is promising for gas exploration.