Intelligent velocity picking and uncertainty analysis based on the Gaussian mixture model

Intelligent velocity picking considering an expert experience based on the Chan–Vese model and mean-shift clustering

“…After obtaining a CMP velocity spectrum, the processor first analyzes the trend and range of velocity for that CMP using empirical and geological knowledge; then, the energy clusters of effective reflected waves within the correct trend are identified with the naked eye and cluster centers are picked up as the velocity for that location. The entire picking is the process by which processors translate geophysical and geological theories into the geometry of energy clusters on the velocity spectrum (Wang et al, 2022). Our method follows this process to achieve automatic velocity picking.…”

“…In addition, Wang et al (2021b) proposed an approach based on adaptive threshold-constrained K-means; this method can improve the identification of the energy cluster of velocity which was weak on the velocity spectrum. Wang et al (2022) also suggested a Gaussian mixture clustering method to achieve automatic velocity picking, as an extreme case of the Gaussian mixture model, and K-means is difficult to characterize velocity energy clusters with low focus ability, while the Gaussian mixture model can accurately fit and provide uncertainty analysis at the same time. These unsupervised clustering algorithms, however, only consider the ability to identify energy clusters of the velocity, without taking into account the complexity of the actual seismic data and expert experiences, resulting in the velocity picking methods affected by the multiples and some other noises on the velocity spectrum.…”