Automatic detection and imaging of diffraction points using pattern recognition

作者：de Figueiredo J J S^*; Oliveira F; Esmi E; Freitas L; Schleicher J; Novais A; Sussner P; Green S

来源：Geophysical Prospecting, 2013, 61: 368-379.

DOI：10.1111/j.1365-2478.2012.01123.x

• Diffraction points
• kNN classifier
• Pattern recognition

Hydrocarbon reservoirs are generally located beneath complex geological structures. Frequently, such areas contain seismic diffractors that carry detailed structure information in the order of the seismic wavelength. Therefore, the development of computational facilities capable of detecting diffractor points with a good resolution is desirable but has been a challenge in the area of seismic processing. In this work, we present a method for the detection of diffraction points in the common-offset-gather domain. The method applies a two-class k nearest neighbours (kNN) pattern recognition technique to amplitudes along diffraction traveltime curves to distinguish between diffractions and reflections or noise. While the method, in principle, requires knowledge of the migration velocity field, it is very robust with respect to an erroneous model. Numerical examples using synthetic seismic and field ground-penetrating-radar (GPR) data demonstrate the feasibility of the technique and show its usefulness for automatically mapping diffraction points in a seismic section. In our applications, the method was able to detect all diffractions present in the data and did not produce any false positives.

• 出版日期2013-6