13 August 2018

Paolo Dell’Aversana*1, Gianluca Gabbriellini1, Gabriela Carrasquero1, Alfonso Amendola1, Alfonso Iunio Marini1 1Eni S.p.A. Upstream and Technical Services As it happens for natural intelligence, also artificial intelligence can be improved if it is able to analyze and interpret multimodal information. In this e-lecture, we show that training a computer with multimodal data, increases the possibility of seismic facies classification through machine learning algorithms. We use a new class of attributes in geophysics, representing ‘musical’ properties implicitly included in the data. Together with traditional geophysical attributes, our multimedia machine learning system uses also melodic, harmonic and rhythmic patterns extracted from the data. All these new features show high classification power. In particular, they allow distinguishing low-gas from high-gas saturated sands, as showed in the real example discussed in this lecture.

View more E-Lectures   Read Paper in EarthDoc    Join EAGE today!

Access to recent EarthDoc material is free of charge for EAGE members.