Automatization of Seismic Migrated Data Segmentation by Implementing Machine Learning Techniques

The seismic survey is one the most important steps in oil and gas exploration. Reconstructing and interpretation of raw data is a very complex engineering problem, requiring a lot of time to complete. Modern computer technologies allow us to perform such operations automatically by implementing methods of machine vision and deep learning techniques. The main condition for such kind of algorithms is the availability of data for training. At this step, we are facing several problems. Oil and gas companies do not publish their data as an open source and it leads to the lack of training data what makes usage of deep learning impossible. In this work, we suggest using machine learning techniques in order to consider the problem of segmentation as the problem of classification. In this case, it is possible to avoid the usage of deep learning and use separate parts of available seismic data for training. According to the results of the work, several software modules for automatic horizons picking were created. In addition, it was conducted the analysis of available methods of machine learning and realized the visualization of the results.